Determination of the Proton’s Weak Charge via Parity Violating Electron Scattering Joshua Russell Hoskins Williamsburg, Virginia Master of Science, College of William and Mary, 2007 Bachelor of Arts, Eastern Kentucky University, 2006 A Dissertation presented to the Graduate Faculty of the College of William and Mary in Candidacy for the Degree of Doctor of Philosophy Department of Physics The College of William and Mary August 2015
202
Embed
Physics | William & Mary - Determination of the …physics.wm.edu/~armd/JoshHoskinsThesisFinal.pdfDetermination of the Proton’s Weak Charge via Parity Violating Electron Scattering
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Determination of the Proton’s Weak Charge via Parity Violating ElectronScattering
Joshua Russell Hoskins
Williamsburg, Virginia
Master of Science, College of William and Mary, 2007Bachelor of Arts, Eastern Kentucky University, 2006
A Dissertation presented to the Graduate Facultyof the College of William and Mary in Candidacy for the Degree of
The road to this point, completing my studies, writing this dissertation, has been along one and I would be remiss to say that I have gotten here on my own merits;we truly are the sum of our parts. Throughout my studies, I have had supportivecolleagues, friends, and family. First, I should thank my family. My interest inscience and discovery was fostered early with my father teaching a fifth graderabout chemistry, physics, and the occasional engineering project such as buildingmy own stereo speaker from household materials or building a potato cannon.Though I wasn’t always a focused student, I was lucky enough to have theencouragement and support that allowed me to find the interests that started thejourney I am now finishing. Without the support received from my mother andfather, getting here would not have been possible.
Throughout my undergraduate and graduate studies I was lucky enough meet andwork with a number of wonderful people. From Eastern Kentucky University,where I did my undergraduate studies, I would like to thank Dr. Garett Yoder, Dr.Marco Ciocca, and Dr. Jerry Cook for their help and guidance. A special thanksgoes to Dr. Christopher Kulp who I was lucky enough to work with during myundergraduate research, and who is the reason I am now graduating from theCollege of William and Mary. During my graduate career at William and Mary Iwas lucky enough to have two excellent scientific mentors; Dr. David Armstrongand the late Dr. Mike Finn. Without the latter I would not have gotten into thefield of experimental particle physics. He was a mentor and friend who is dearlymissed by many; myself included. I would also like to thank David Armstrong fortaking up the mantle of advising me and seeing me through to the end. Withouthis insight and mentoring I could not have come this far; it was a pleasure andhonor working with him.
With any scientific endeavour, there are often many people, working countlesshours to do great things. The Qweak experiment, being no exception, was madepossible by a large number of talented and hard working individuals both insideand out of academia. I would like to thank all of the Qweak collaboration for theirhard work and dedication for making this measurement possible. Special mentionincludes Roger Carlini, Greg Smith, Paul King, Mark Pitt, Kent Paschke, StephenWood, Wouter Deconinck, and Jeong Han Lee. From Jefferson laboratory I wouldlike to give special thanks to David Mack for his countless hours of discussion andmentoring; if knowledge imparted to graduate students was money you would be a
v
rich man. I would also like to thank Brad Sawatzky for his insight, friendlydiscussions, debugging, and coffee. Lastly, I would like to thank Dave Gaskell forhelping me throughout my time at the laboratory; without Dave I would havegotten nothing done in Hall C.
In addition to the scientist at Jefferson Lab, I would like to recognize staff andengineers that made the experiment not only a success but possible at all. Theseinclude Chris Cuevas and the Fast Electronics group, the Polarized Target group,Walter Kellner, Andy Kenyon, Paulo Medeiros, and all of the Hall C technicalstaff. Without your work none of the work presented in this dissertation wouldhave been possible.
As with any experiment there is often an army of graduate students workingbehind the scenes to make things happen. I would like to thank you all for yourhard work and dedication. Special mention goes to John Leacock, KatherineMyers, Amendra Narayan, Scott MacEwan, Josh Magee, Donald Jones, JohnLeckey, Nuruzzaman, Rakitha Beminiwattha, Adesh Subedi, and BuddiniWaidyawansa. It is only through your work that we were able to finish thisexperiment. It was a pleasure and honor to work with you all.
Lastly, I must thank my wonderful wife Kerry. She has been an unyielding sourceof love, support, reassurance, coffee, and proofreading in my life over the last nineyears. Whether it was small things around the house when I was busy at work,bouts of doubt and depression, or helping with the latest piece of mechanicaldesign, I am truly lucky to have met someone like her. I am here because you werethere.
vi
I dedicate this thesis to my parents and my wife Kerry.
vii
LIST OF TABLES
2.1 The vector and axial couplings interacting with Z0 are shown for each
A.1 The survey team used tooling balls that are attached in static lo-
cations on the VDCs to determine position in the frame of the lab.
Above shows the positions and designation of each tooling ball. . . . 169
xvii
DETERMINATION OF THE PROTON’S WEAK CHARGE VIA PARITY
VIOLATING ELECTRON SCATTERING
CHAPTER 1
Introduction
1.1 Fundamental Symmetries of the Standard Model
1.1.1 Fundamental Symmetries
The concept of fundamental symmetries has played a major role in the devel-
opment of the Standard Model of particle physics and has been one of the most
ubiquitous elements in the formulation of physics in the 20th century. Conservation
laws in physics are related directly to the invariance of a physical system under
a transformation. In 1918, German mathematician Emmy Noether published her
theorem[15] which states that in the case of a system having a continuous symmetry,
there will be corresponding quantities whose values are conserved, i.e. symmetries
which lead to the invariance of the action under transformation lead to conserved
quantities. This is an important result because it gives insight into conservation
laws intrinsic to physical systems, as well as providing a practical calculation tool
for conserved quantities. Symmetries can be either discrete or continuous. An exam-
ple of a discrete symmetry is reflective symmetry or parity. In the case of continuous
2
symmetry, familiar examples from classical mechanics would be the invariance of a
system under rotation which leads to conservation of angular momentum or transla-
tional invariance in time which gives conservation of energy. In quantum mechanics,
conservation principles are tied to the commutation an of operator with the Hamil-
tonian. For example, consider a general operator in the Hamiltonian picture,
dOdt
= i[O, H]. (1.1)
Here the commutator with the Hamiltonian describes the time evolution of the
operator. In the event that the operator, O, commutes with the Hamiltonian, the
physical observable associated with that operator does not change with time and is
therefore conserved. In Field Theories, symmetries are defined by transformations
on a physical system that leave the action unchanged. The action is given in terms
of the Lagrangian density by
S =
∫d4xL(φ, ∂µφ) (1.2)
This can be simplified by looking at transformations that leave the Lagrangian
density invariant. Transformations can be global or local, each of which affects the
Lagrangian in different ways. As an example, consider the Dirac Lagrangian of a
charged spin 1/2 particle of mass m,
L = ψ(iγµ∂µ −m)ψ. (1.3)
A global transformation can be represented as a change of phase,
ψ → ψ′ = eiαψ. (1.4)
3
It is clear that replacing ψ → ψ′ in Eq. 1.3 leaves the Lagrangian unchanged
because the transformation affects all points in space-time equally. A local, or gauge,
transformation is space-time dependent, however, and must be handled differently.
A gauge transformation can be represented as
ψ → ψ′ = eiα(x)ψ. (1.5)
Substitution of Eq. 1.5 into Eq. 1.3 finds the Lagrangian is not, at least a priori,
invariant under a local gauge transformation. Gauge invariance can be imposed by
introducing a gauge field with a transformation property such that the extra term
is cancelled. We define the covariant derivative to be
Dµ ≡ ∂µ − igAµ (1.6)
where Aµ transforms as
Aµ → Aµ +1
g∂µα. (1.7)
Thus, requiring gauge invariance introduces a vector field, Aµ, that couples directly
to the charged particles described by the Dirac Lagrangian. In fact, choosing the
coupling constant, g, to represent the electric charge, e, we see this new vector
field represents the photon field. Thus, requiring gauge invariance introduces a new
vector field that couples to each particle in the theory and becomes the force carrier
for the interaction. These new particles are called vector bosons. As a note, the
addition of Aµ potentially adds both a kinetic, FµνFµν , and a mass term, mAµA
µ,
to the Dirac Lagrangian. The latter of these violates gauge invariance therefore the
mass term cannot exist, the photon field is massless and infinite in range. Vector
4
bosons are discussed in more detail in subsequent chapters.
As mentioned above, requiring gauge invariance introduces new vector bosons
into the theory which in the case of the photon field are massless. This becomes a
problem, however, when applying gauge invariance to the weak interactions, where
the charge carrier vector bosons (Z, W±) masses have been measured to be on the
order of 100 GeV. The solution is spontaneous symmetry breaking. Spontaneous
symmetry breaking describes a situation where the underlying laws at low energies
have symmetries which are hidden. This mechanism produces the charged vector
bosons (Z, W±) as well as giving mass to the fermions. This is know as the Higgs
Mechanism and it plays a crucial part in our understanding of the Standard Model.
1.1.2 Standard Model Overview
At present the interactions and constituents that underlay the observable uni-
verse have been reduced to a handful of physical laws defined by the Standard Model
(SM). The SM is the theory of the electromagnetic, weak, and strong forces, as well
as the particles that make up the building blocks of matter, and how these forces
mediate the subatomic world. The SM in its current form was mostly finished in
the early 1970’s starting with the confirmation that the proton was made up of
smaller constituents[16, 17]; at the time these were called partons, however they
were later identified as the up and down quarks. The early theoretical development
of the SM started with Glashow’s 1961 combination[18] of the electromagnetic and
weak interactions to form the SU(2)L × U(1)Y gauge group, which created elec-
troweak theory. After the addition of the Higgs[19] by Weinberg and Salam[20, 21]
in 1967, the electroweak theory in its current manifestation was mostly complete.
The addition of the quark model proposed by Gell-Man[22] and Zweig[23] added
5
SU(3)C color symmetry defining the current form of the SM as operating under the
U(1)Y × SU(2)L × SU(3)C gauge symmetry.
Since the coming together of electroweak theory and the quark model defined
the current SM, experimental tests have been used to systematically verify its valid-
ity. In the mid-1970’s the discovery of neutral-weak currents generated via Z boson
exchange[24, 25] helped confirm electroweak unification for which Glashow, Salam,
and Weinberg later shared the 1979 Nobel prize. The Prescott experiment[26] per-
formed at the Stanford Linear Accelerator, in the late 1970’s, saw the experimental
measurement of the SM predicted parity-violating asymmetry in inelastic electron-
Deuteron scattering; this was an important step in the wide acceptance of the SM.
The construction of the proton-antiproton collider at CERN in the late seventies
opened the door to the production of particles two orders of magnitude above the
proton mass. This lead to the experimental discovery of of the W±[27, 28] and
Z[29, 30] and was of fundamental importance in validating a crucial element of the
SM; this discovery earned Carlo Rubbia and Simon van de Meer the 1984 Nobel
Prize. Over the course of the last 30 years experimental verification of the SM has
continually mounted, and with the recent discovery of what is likely to be the Higgs
[31, 32] at CERN, the final building block of the SM has been experimentally ob-
served and verified. To date, the SM has been successful in explaining nearly all
of high-energy experimental data and has fostered a rich program of experimen-
tal programs testing fundamental theories and searching for evidence of possible
extensions.
The SM describes the laws governing the building blocks of matter and their
interactions (Fig. 1.1). The SM describes all matter as being made up of 12 spin
1/2 particles known as fermions and 4 vector bosons that govern their interactions;
a spin-0 scalar boson, the Higgs, gives the particles mass. The fermions are divided
6
FIG. 1.1: The Standard Model of Particle Physics as presently determined. Fermions(1/2 integer spin) are divided into three generations with similar properties andincreasing mass. Fermions are divided into leptons (do not interact via the strongcoupling) and quarks. Force carriers are represented by the integer spin bosons tothe right. Together these make up the Standard Model and have been extremelysuccessful in explaining experimental data over the last 50 years[1].
into quarks and leptons. There are 6 leptons, three of which carry a negative
integral charge: electron(e), muon(µ), tau(τ). Neutrinos are neutral in charge and
designated as: electron neutrino(νe), muon neutrino(νµ), tau neutrino(ντ ). Each
lepton has the intrinsic properties of mass, charge, and spin. The quarks are defined
in a similar manner with the first generation being defined as up(u) and down(d), the
second being charm(c) and strange(s), and the third being top(t) and bottom(b).
Each quark has an associated intrinsic mass, charge, spin, and color charge. Each
generation of lepton is increasingly more massive, with the first generation being
the lightest; the heavier generations are in fact not stable and quickly decay with
short lifetimes. Interestingly, because the first generation of charged fermions do not
7
decay, they make up all baryonic (formed from a bound state of 3 quarks) matter
in the universe.
The force carriers of the SM define how particles interact with each other. In
the SM, these force carriers are the spin-1 vector bosons. The charged interaction
or the electromagnetic force between both leptons and quarks is mediated by the
photon. Classically, this coupling is what we think of as the electric charge; with the
unification of the electromagnetic and weak gauge groups into electroweak theory,
charge actually becomes a function of both weak isospin and weak hypercharge.
This is discussed more in section 1.2.2. The weak charge is mediated via three
massive vector bosons: Z, W+, and W−. Due to their mass, each force carrier’s
interaction range is short, as opposed to that of the photon, which is massless and
has an infinite range. The neutrinos, being neutral particles with no charge, interact
only via the weak charge making them difficult to detect. The strong force, which is
responsible for quark binding within the nucleus, is mediated by the eight massless
gluons via coupling to color charge. Leptons, being devoid of color charge, do not
interact via the strong force; this is what separates the fermions from the leptons.
Interestingly, unlike leptons which can be observed as free particles, quarks are
only found in bound states. Each quark has a color charge (Red, Blue, Green),
and interacts under the strong, weak, and electromagnetic interactions. The bound
states of quarks come in colorless combinations of three (baryons), pairs (mesons),
and possibly other combinations. The fact that quarks are only found in bound
states is a demonstration of color confinement; confinement also has the interesting
property that, unlike the electroweak force, the strength of the coupling increases
as the distance between quarks increases.
Despite the obvious success of the SM in the explanation of the interactions and
the constituents making up the universe, it has a number of significant failures. One
8
of the most significant shortcomings is that there is no way of reliably describing
the canonical theory of gravitation, General Relativity, in terms of modern quantum
field theory. Other problems such as a lack of explanation of the matter/anti-matter
asymmetry in the universe and lack of a proper dark matter candidate particle have
lead to the exploration of Beyond the Standard Model(BTSM) physics, as well as
experimental programs centred around testing SM observables in an effort to find
discrepancies; the discrepancies could give important insight into the nature of new
physics.
1.2 Electroweak Theory
1.2.1 Discrete Symmetry and Spin
Mentioned briefly in the beginning of the chapter (Subsec. 1.1.1) was the idea of
discrete symmetries which play an important role in the Standard Model; specifically
parity. Parity describes the way a system behaves under a spatial transformation,
(x, y, z) → (-x, -y, -z). This is often referred to as mirror symmetry as it is a
reflection through the origin and manifests in a similar way to looking in a mirror.
Originally, parity was thought to be a conserved quantity in the SM; experimental
evidence in both the electromagnetic and strong interactions indicated that parity
was conserved. In 1956 T. D. Lee and C. N. Yang, while looking at the question
of parity conservation in β, hyperon, and meson decays, pointed out that there was
no a priori reason why parity should be conserved in the weak interactions. In
fact there was no experimental evidence for (or against) parity violation in the weak
interactions [33]. Subsequently, at the National Bureau of Standards, an experiment
using β-decay in 60Co nuclei was carried out by C.S. Wu and colleagues [34] to test
9
the theory set forth in Yang and Lee’s paper. The idea of the experiment was
to look at the decay distribution of 60Co nuclei in a magnetic field. If the weak
interactions, which mediated the decay process, conserved parity then the measured
electron emission should be the same for both 60Co spin aligned and anti-aligned with
the magnetic field. What was found was a measurable asymmetry in the detected
emission indicating a preferred direction in the decay process. This provided the
first experimental evidence of parity violation in the weak interactions and lead to
a 1957 Nobel prize for Yang and Lee.
An important property to consider, especially in the discuss parity and the weak
interactions, is the way in which “handedness” behaves under parity transformation.
The spin of a particle can be used to define the “handedness” of a particle and is
often referred to as the helicity or in the massless, relativistic limit, the chirality.
Helicity describes the orientation of the spin vector with respect to the momentum
of the particle. For a right-handed particle, the momentum and the spin would be in
the same direction, and for left-handed would be opposite of each other. A particle’s
chirality is more subtle, and is defined by the chiral “projection operator” shown in
Eq. 1.8 and 1.9.
uL(p) =1
2(1− γ5)u(p) (1.8)
uR(p) =1
2(1 + γ5)u(p) (1.9)
Here the projection operator acts on the particle spinor returning a left(right)-
handed chiral particle. A key difference between helicity and chirality can be un-
derstood considering a Lorentz boost. For a massive particle it is possible to boost
a left-handed particle such that the helicity is reversed. This leads to the situa-
tion where the particle doesn’t interact the same in all reference frames. This is
not true of the chirality which is an intrinsic property of the particle and invariant
10
under change of reference frame. As mentioned above, in the limit where the par-
ticle is massless and relativistic, helicity and chirality become the same. In further
discussion contained in this document, only the case of massless, ultra-relativistic
electrons will be considered.
1.2.2 Electroweak Unification
As mentioned in Sec.1.1.2, the vector bosons, W± and Z, act as the force
carriers of the weak force. The vector bosons also interact under the electromag-
netic force. The fact that the force carriers have a charge associated with both
the electromagnetic force as well as the weak force hints at the unification of the
electromagnetic and weak forces. The unification of both the weak and electromag-
netic interactions into a single theoretical framework, in which they would appear
as different manifestations of a single theory, was the goal of Glashow’s early work
[18]. The addition of later work by Weinberg and Salam[20, 21] lead to the eventual
development of the current electroweak theory. In the following section I will briefly
discuss some of the underlying ideas of electroweak unification.
The weak and electromagnetic theories are mathematically unified under the
SU(2)L × U(1)Y gauge group. Recalling the previous discussion of gauge invari-
ance of the Lagrangian (Sec. 1.1.1), we can define the covariant derivative for the
electroweak Lagrangian to be,
Dµψ = (∂µ +ig
2τ iW i
µ +ig′
2Y Bµ)ψ. (1.10)
Here Wµ represents the charged vector boson triplet required for gauge invariance
under SU(2)L, and τ i, known as the Pauli spin matrices, are generators of the weak
isospin symmetry. The vector boson singlet is given by Bµ, and the generator of the
11
U(1)Y symmetry is given by the hypercharge, Y. The couplings for the electroweak
theory are given for SU(2)L and U(1)Y as g and g′ respectively. Under the unified
theory the electric charge is redefined in terms of weak isospin and the new U(1)
symmetry hypercharge as
Q = T 3 +Y
2. (1.11)
In a sense this unifies the electromagnetic and weak interaction, albeit with two inde-
pendent coupling strengths. Using the covariant derivative (1.10), the Lagrangian
for the electroweak theory can be constructed; mass terms of the form mψψ are
excluded due to failing gauge invariance. The fermion masses in the electroweak
theory are generated by spontaneous symmetry breaking via the Higgs Mechanism
in which the degrees of freedom of the scalar Higgs field are “absorbed” by the
massive gauge bosons. In short, a set of complex scalar fields, the Higgs, can be
introduced, resulting in a breaking of the SU(2) global symmetry. As a result, the
theory produces three massive vector bosons, W± and Z, and one massless boson,
the photon(γ), as well as giving mass to the fermions. The Lagrangian describing
the new scalar fields is given by
L = (DµΦ)†(DµΦ)− µ2Φ†Φ +λ2
4(Φ†Φ)2 (1.12)
where
Φ =
φ+
φ
(1.13)
is the complex scalar Higgs doublet. The choice of µ2 here matters. If µ2 > 0 is
chosen the theory returns the standard QED theory with a massless photon and a
charged scalar. Choosing µ2 < 0 however results in the “mexican hat” potential
12
which has a non-zero vacuum expectation value (〈Φ〉 6= 0), i.e. the symmetry is
broken. Minimizing the potential in the above Lagrangian and choosing an arbitrary
ground state breaks the SU(2) symmetry. This choice is important because the
minima lie on a circle of radius µ2/λ2 and therefore there are an infinite number
of solutions. Expanding 1.12 above allows for identification of the charged vector
bosons as
W± =1√2
(W 1 ∓ iW 2) (1.14)
and the neutral vector bosons as
Z0µ = cos θwW
3µ − sin θwBµ (1.15)
Aµ = sin θwW3µ + cos θwBµ. (1.16)
Here the physical vector bosons we see in experiment, and which gain mass through
symmetry breaking, are defined as a mixing between the neutral bosons W 3µ and Bµ.
The parameter θw is the “weak mixing angle” and describes the mixing between the
neutral vector boson couplings. The masses acquired by the vector bosons through
the Higgs mechanism are give by:
m(W±µ ) =
1
4g2ν2 (1.17a)
m(Z0µ) =
1
4(g2 + g′2)ν2 (1.17b)
m(Aµ) = 0 (1.17c)
Following this further, the vector boson masses can be related as
cos θw =mW
mZ
. (1.18)
13
This answers one of the central questions in electroweak physics; why are the masses
of the W± and Z0 different? It is also important to note that under unification the
couplings of electroweak theory are not independent. In the interest of foreshadow-
ing, g and g′ can be related as
sin2 θW =g′2
g2 + g′2. (1.19)
The term sin2 θW is a center piece of the work in this thesis and will be discussed at
length in subsequent chapters.
Lastly, the fermions gain mass via interaction with the Higgs. Considering the
general SU(2) × U(1) Yukawa coupling of a scalar particle interacting with fermions
along with the Higgs doublet defined by
Φ =
0
ν + h(x)
(1.20)
where h(x) is a small perturbation about the ground state. The interaction La-
grangian for the Higgs coupling to the fermions is given by
LY U = g(e)ψLΦeR + h.c. (1.21)
where ψL describes the SU(2) electron doublet and g(e) represents the Yukawa cou-
pling. For simplicity I have only included minimal terms in the interaction La-
grangian; this can be expanded to include all fermions. Expanding upon this, the
fermions masses can be seen as coefficients of terms that are quadratic in the fields.
It is interesting to note that the actual masses of the fermions are not predicted
by the theory and are only input parameters. The above work only shows how the
14
masses are formed under the theory.
15
CHAPTER 2
The Qweak Experiment
2.1 Experimental Motivation
The Standard Model (SM), while being successful in describing the fundamen-
tal interactions found in nature, is thought to be an effective low-energy theory of
a more fundamental underlying physics. There are two complementary methods of
searching for new physics: that of high energy experiments which strive to excite
matter into new forms, and that of precision experiments which aim to measure ob-
servables in the SM that are precisely predicted. Historically, precision experiments
have been crucial in studying the structure of the nucleon and understanding the
electroweak interaction. The weak charge of the proton, Qpw = 1 − 4 sin2 θw (tree
level), which is the neutral-weak analog of the proton’s electric charge[35], is both
precisely predicted and moderately suppressed in the SM. Measurements made at
the Z-pole have done an impressive job providing constraints and verifying predic-
tions of the sin2 θw at high energy. A lesser studied area, which has great potential,
is measurement of sin2 θw at low Q2. The SM predicts a shift of ∆ sin2 θw = 0.007 at
16
low Q2 with respect to the Z-pole value. This shift comes about due to the energy
dependence of the weak coupling; as Q2 goes higher, radiative corrections shift the
value from the measured value at the pole. Significant deviation from the theo-
retically predicted value would be a strong indication of new physics, while precise
agreement would provide an important stand-alone confirmation of the SM. Thus,
a measurement of Qpw provides an excellent candidate for indirect searches of new
physics; specifically parity-violating (PV) physics in the coupling between electrons
and light quarks.
The Qweak experiment, which ran at the Thomas Jefferson National Accelerator
Facility (JLab) from November 2010 to May 2012, provides the first direct determi-
nation of Qpw via a precise measurement of the PV asymmetry in ~ep scattering at
low momentum transfer (Q2 ≈ 0.025 GeV2). The choice of low momentum transfer
and the use of parity-violating electron scattering (PVES) world data, which helps
to constrain errors, allows for a theoretically clean extraction of Qpw. First results
were recently released using only 4% of our full data set[36]. Precise measurement of
the weak-charge, which can be written in terms of the vector quark weak charges as
Qpw = −2(2C1u + C1d), also provides an important complement to precision atomic
parity-violation (APV) experiments. APV experiments on 133Cs[37] provide access
to a different linear combination of the vector quark weak charges that can be used
to separate and determine C1u and C1d. The following chapter aims to lay out
the basic theoretical framework and possible implication of the Qweak experiment’s
measurement.
17
2.2 Parity-Violating e+p Scattering
2.2.1 Neutral-Weak Interaction
The neutral-weak interaction is of fundamental importance to PVES experi-
ments. As explained in Sec. 1.2, the interactions in the SM are mediated by the
transfer of gauge bosons. The neutral-weak interactions describe interaction via γ
or Z0 bosons, and are considered neutral because the interaction does not affect the
charge of the incoming and outgoing particles. The neutral-weak current for both
the γ and Z0 are given by
J NCµ =
g
cos θW[J 3
µ − sin2 θWJ EMµ ] (2.1)
J EMµ = ψQγµψ. (2.2)
Here J EM is the electromagnetic current and J 3 is the current associated with
the third component of the SU(2)L isospin triplet. The mixing intrinsic to the
electroweak interaction, explained in the previous chapter, is manifest in the make
up of the neutral-weak current being in terms of J 3 (SU(2)L) and J EM (U(1)Y ).
For simplicity, only the electron fields will be considered. Equations 2.1 and 2.2 can
be expanded as
J NCµ =
g
cos θW[ψ
eγµ(
1
2(1− γ5) +Q sin2 θW )ψe]. (2.3)
It is instructive to write Eq. 2.3 in terms of left and right handed fields
J NCµ =
g
cos θW[ψ
e
Lγµ(−1
2+Q sin2 θW )ψeL + ψ
e
Rγµ(Q sin2θW
)ψeR], (2.4)
18
where the substitutions,
ψL =1
2(1− γ5)ψ (2.5)
ψR =1
2(1 + γ5)ψ (2.6)
have been made. Here we can see one of the most important intrinsic properties
of the neutral-weak current; the property which is the basis of PVES experiments.
The neutral-weak force interacts differently between left-handed and right-handed
particles, and it therefore violates parity. It is important to note that neither the
vector (γµ) nor the axial (γ5) part of 2.3 violates parity. The vector part flips sign
under parity while the axial part does not; considering the square of the scattering
amplitude neither pure vector nor pure axial would violate parity. It is the inter-
ference of the V-A coupling that is the basis of the neutral-weak current violating
parity. The general form of Eq. 2.3 is given in terms of the vector and axial parts
as
J NCµ =
g
cos θWψeγµ(cfV + cfAγ
5)ψe, (2.7)
where ceV and ceA and the vector and axial couplings respectively and f is the fermion
flavor. Comparing this with Eq. 2.3 above the couplings can be identified as ceV =
−12
+ 2 sin2θW
and ceA = −12.
The results for the cV and cA couplings depend on what flavour of fermion is
interacting with the Z0. The results for each flavour as well as the electric charge
are given in Table. 2.1.
2.2.2 Low Momentum Electron Scattering
At low momentum, the tree-level scattering amplitude of electron-proton scat-
tering is mediated by either the γ or Z0. The tree-level scattering amplitude in terms
19
Fermion Flavor Qf cfA cfVνe, νµ, ντ 0 1
212
e, µ, τ -1 -12
-12
+ 2 sin2 θWu, c, t 2
312
12
+ 43
sin2 θWd, s, b -1
3-1
2-1
2+ 2
3sin2 θW
TABLE 2.1: The vector and axial couplings interacting with Z0 are shown for eachflavor of lepton.
FIG. 2.1: Tree level diagrams for ep scattering in the case of the electromagneticand neutral-weak interactions.
of currents is
Mγ,Z = J e,γµ J µ
p,γ + J e,pµ (V µ
p,Z − Aµp,Z). (2.8)
The electron currents for both the electromagnetic and neutral weak interactions
were derived in the previous section (Sec. 2.2.1) and are shown in Eq. 2.9a and Eq.
2.9b in terms of the vector and axial couplings.
J e,γµ = −euγµu (2.9a)
J e,Zµ = uγµ(gV + gAγ5)u (2.9b)
The interaction at the proton’s vertex is more complicated because, unlike the elec-
tron, the proton has an internal structure and so the amplitude involves complicated
interactions of the force carrier with the valence and sea quarks. Instead we rep-
resent the current at the proton’s vertex in terms of structure functions. These
structure functions, or form factors, give the properties of the particle interactions
20
at a given momentum transfer without having to include the underlying physics
directly. This is important because often the calculations of the underlying physics
cannot be done at present. The proton’s electromagnetic current can be written as
J µp,γ = u[γµFN
1 (Q2) + σµνqν
2MFN
2 (Q2)]u (2.10)
where F1 and F2 are the Dirac and Pauli form factors respectively. These form
factors contain complete information about the elastic electric and magnetic inter-
action of the nucleon as a function of the four momentum transfer Q2. The Sachs
electric and magnetic form factors are given in terms of the Dirac and Pauli form
factors as
GE(Q2) = F1(Q2)− Q2
4M2F2(Q2) (2.11a)
GM(Q2) = F1(Q2) + F2(Q2). (2.11b)
Some intuition about the Sachs form factors can be gained by considering scatter-
ing in the Breit frame, i.e. the frame in which the initial and final state nucleon’s
momenta have the same magnitude. In this reference frame the Sachs electric form
factor GE can be interpreted as the Fourier transform of the charge distribution; the
magnetic form factor gives the Fourier transform of the magnetization distribution
in the proton, which at Q2 → 0 give the anomalous magnetic moment, µp.
In addition to the electromagnetic coupling of the proton, the neutral-weak
coupling must also be considered. The vector (Vp,Z) and axial (Ap,Z) combinations
21
of the neutral-weak form factors are
V µp,Z = u[γµFZ
1 (Q2) + iσµνqν
2MFZ
2 (Q2)]u (2.12a)
Aµp,Z = u[γµγ5GZA(Q2)]u (2.12b)
The neutral-weak vector form factors are given by FZ1 and FZ
2 and the axial contri-
bution is given in terms of the axial form factor (GZA). The neutral-weak form factors
are less experimentally known compared to the electromagnetic form factors. It is
helpful, given the internal structure of the nucleon, to write the neutral-weak form
factors in terms of the sum of their quark contributions. Neglecting contributions
from the more massive quarks, the form factors can be written as
Gγ,pE,M =
2
3Gu
E,M −1
3Gd
E,M −1
3Gs
E,M (2.13a)
GZ,pE,M =
1
2(1− 1
8sin2 θw)Gu
E,M −1
2(1− 4
3sin2 θw)Gd
E,M −1
2(1− 4
3sin2 θw)Gs
E,M
(2.13b)
GZ,pA = Gs
A +GdA −Gu
A, (2.13c)
where GiE,M is the ith quark form factor. The nucleon flavor forms factors above
carry contributions from both the valence and sea quarks; the exception is the
strange form factor which only a contribution from the quark sea given the fact that
the nucleon has no net strangeness. At low energies the proton and neutron can
be thought of as different manifestations of the same particle; this defines isospin
22
symmetry. Using isospin symmetry,
Gu,pE,M = Gd,n
E,M (2.14a)
Gu,nE,M = Gd,p
E,M (2.14b)
to rewrite 2.13, we can define the neutral-weak form factor in terms of the electro-
magnetic form factors of the proton and neutron as
GZ,pE,M = (1− 4 sin2 θw)Gγ,p
E,M −Gγ,nE,M . (2.15)
2.2.3 Physics Asymmetry
The elastic process of an electron scattering from a proton contains contribu-
tions from both the electromagnetic and neutral-weak interactions. At low momen-
tum transfer, the tree-level total invariant amplitude scattering can be expressed as
the square of the sum of the diagrams for γ and Z0 exchange,
|Mep|2 = |Mγ +MZ |2 (2.16a)
= |Mγ|2 + |MZ |2 + 2<(M∗γMZ). (2.16b)
Because Mγ MZ , the electromagnetic amplitude dominates in terms of contri-
bution, however because the neutral weak current doesn’t conserve parity it can
be isolated by defining the difference between the scattering amplitudes, or the
parity-violating asymmetry. The PV asymmetry (Aep) for longitudinally polarized
electrons scattering from unpolarized protons can be defined as the difference be-
tween the scattering cross section of positive and negative helicity electrons over the
23
total scattering cross section,
Aep =dσL − dσRdσL + dσR
. (2.17)
Aep ≈2Re(M∗
γMZ)
|Mγ|2(2.18)
The crucial piece of the asymmetry lies in the numerator. The asymmetry of M2γ
disappears due to parity conservation, leaving only the M2Z amplitude and the
interference term. Here the interference term dominates and the asymmetry is
approximated by
Aep ≈2ReM∗
γ(MZ,L −MZ,R)
|Mγ|2(2.19)
The fact that the weak interactions violate parity allows us to isolate the neutral
weak contribution which would otherwise be lost in the electromagnetic scattering
amplitude. At tree level this can be expressed in terms of the Sachs electromagnetic
form factors and grouped into three pieces as
A =GFQ
2
2√
2πα[AE + AM + AA] (2.20)
where the electromagnetic, magnetic, and axial asymmetries define groups of Sach’s
form factors,
AE =εGγ
EGZE
ε(GγE)2 + τ(Gγ
M)2(2.21a)
AM =τGγ
MGZM
ε(GγE)2 + τ(Gγ
M)2(2.21b)
AA =12
√τ(1− ε2)(1 + τ)Gγ
MGZA
ε(GγE)2 + τ(Gγ
M)2(2.21c)
24
Both AE and AM terms arise due to the axial electron current coupling to the vector
current of the proton, whereas the axial term, AA, arises from the vector coupling
of the Z0 to the electron and the axial coupling to the nucleon. Expanding Eq. 2.20
in terms of the form factors,
Aep = A0
[εGγ
EGZE + τGγ
MGZM − (1− 4 sin2 θW )ε′Gγ
MGZA
ε(GγE)2 + τ(Gγ
M)2
](2.22)
where
A0 =−GFQ
2
4πα√
2, ε =
1
1 + 2(1 + τ) tan2 θ2
, and ε′ =√τ(1 + τ)(1− ε2) (2.23)
are kinematic quantities, GF the Fermi constant, sin2 θW the weak mixing angle,
−Q2 the four-momentum transfer squared, α the fine structure constant, τ =
Q2/4M2, M the proton mass, and θ the laboratory electron scattering angle. In
the forward angle and low momentum limit this can be rewritten in a more simple
form. Considering the kinematic variables in the limit ofQ2 → 0 it is straightforward
to see that ε→ 1, ε′ → 0, and τ → 0. This reduces Eq. 2.22 to
Aep = A0
[GZ
E
GγE
]. (2.24)
Replacing GZE above with Eq. 2.15 and grouping next-to-leading order terms into
the hadronic term B(Q2, θ) the asymmetry can be written simply as
Aep/A0 = Qpw +Q2B(Q2, θ). (2.25)
At leading order in Q2 the reduced parity-violating asymmetry gives the pro-
ton’s weak charge, Qpw. At next-to-leading order the B(Q2, θ) term contains infor-
25
mation about the electromagnetic, weak, and strange form factors and is relatively
suppressed at low Q2. In choosing the momentum transfer at which the experiment
was run this term was very important. Contributions from B(Q2, θ) can be reduced
by lowering the Q2, however this also reduces the magnitude of Aep, thus our ability
to determine the asymmetry precisely. Setting the momentum transfer to 0.0025
(GeV/c)2 allowed for a precise measurement of Qpwwhile constraining the contri-
bution from B(Q2, θ) to ∼ 30%. The determination of B(Q2, θ) was done using a
global fit of the existing PVES data up to 0.63 (GeV/c)2. This fit as well as details
of the extraction of Qpw is discussed in Section 6.1.
2.2.4 Precision Determination of sin2 θw
FIG. 2.2: The electromagnetic interaction at O(α)(left) and O(α2)(right). Thevacuum polarization screens the bare charge of the electromagnetic interaction atthe vertex.
Thus far in this discussion, the primary focus has been on the leading order con-
tribution to the physics asymmetry, however as either Q2 increases or the precision of
our measurement increases it is important to consider effects of higher-order contri-
butions. Before discussing the implications of higher-order diagrams in the context
of the weak-charge it is first instructive to look at the electromagnetic charge. To
leading order the electromagnetic coupling is given by the fine-structure constant α.
Couplings, in general within the SM, are energy dependent and therefore as Q2 in-
26
creases higher-order diagrams, in powers of the coupling constant, must be included.
The effect on the measured or “effective charge” can be thought of as an electron
in a dielectric. This effect comes about because the vacuum acts as a dielectric;
electron-positron pairs appear out of the vacuum, effectively screening the actual
charge. The way in which the coupling changes as a function of Q2 is called the
running of the coupling.
FIG. 2.3: The running of the weak mixing angle using the MS renormalizationscheme[2]. The width of the curve represents the theoretical uncertainty in thecalculation. The Z-pole value is given at Q2 = MZ .
In the same manner as the electromagnetic coupling, the weak-coupling also
has a dependence on Q2. Calculations of the running of sin2 θw are included in
[2, 38]. The renormalization scheme was chosen to be MS [2] and is shown plotted
as a function of Q in Fig. 2.3. Here the width of the line represents the theoretical
uncertainty of the calculation. The large number of measurements taken at the
Z-pole (Q2 = MZ) anchors the plot and measurements taken away from this Q
provide important tests of the theoretical predictions of the SM. This places special
importance on the understanding of the radiative corrections at the momentum
27
transfer of Qweak. The SM predicts a shift in sin2 θw from the Z-pole value at low
momentum transfer of ∼ 0.007. It is important to note that these corrections are
renormalization scheme dependent. The weak-charge including radiative corrections
where ρNC accounts for one-loop corrections to the gauge boson propagators, which
at tree level is defined to be ρNC ≡M2W/M
2Z cos2 θw. The one-loop corrections come
from the top and bottom quark loops to the gauge boson propagators; there are
contributions from other quark generations but they are negligible. The terms ∆e
and ∆e′ are the photon loop correction to the Z boson exchange vertex and the Z
loop correction to the photon exchange vertex respectively. Diagrams for both ρNC
and ∆e,e′ are shown in 2.4. The final three terms in 2.26 represent box diagrams
FIG. 2.4: The one loop contribution to Qpw from the gauge boson mass renormal-
ization is shown on the left. The γ, Z loop correction to the Z, γ exchange vertex isshown on the right.
describing the exchange of two gauge bosons. The box diagrams for ZZ and WW
are relatively straight-forward to calculate using pQCD due to the propagators of
the W and Z within the box being dominated by high momenta. The γZ diagram
is much more problematic because the photon is dominated by low momentum
exchange which is outside of the useful regime of pQCD. Calculation of the γZ
diagram is discussed in more detail in 2.2.5. The box diagrams can be seen in Fig.
28
2.5. Lastly, sin2 θw(0) is the one-loop definition of sin2 θw evaluated at Q2 = 0.
FIG. 2.5: Box diagrams representing the exchange of two gauge bosons (Cross termsnot shown).
The one-loop definition of sin2 θw here contains contributions from mixing diagrams
in which a Z-boson fluctuates into a photon via a fermion loop, single W loops,
and two W loops containing goldstone bosons. A detailed derivation of the terms
contributing to the running of sin2 θw can be found in [2, 40, 38]. Fig. 2.6 shows
two examples of diagrams that contribute to the running.
FIG. 2.6: Two examples of diagrams contributing to the running of sin2 θw. Theleft diagram shows a Z boson fluctuating into a photon via fermion loop. The rightdiagram shows a single W loop.
2.2.5 γZ0 Box Diagram
In the forward-limit the dominant energy-dependent radiative correction to Eq.
4 comes from the γ − Z box diagram (VγZ(E,Q2)) which arises in the axial-vector
coupling at the electron vertex. This correction has been evaluated using dispersion
29
TABLE 2.2: Recent calculations of VγZ(E,Q2) and its uncertainty at the kinematicsof this measurement.
Reference VγZ(E,Q2) ∆VγZ(E,Q2)
Sibirtsev, et al. [41] 0.0047 +0.0011−0.0004
Rislow, et al. [40] 0.0057 0.0009Gorchtein, et al. [42] 0.0054 0.0020Hall, et al. [43] 0.0056 0.00036
relations [41, 40, 42, 43] independently by several groups, and is summarized in
Table 1. The most recent calculation [43] of VγZ(E,Q2), uses parton distributions
and benchmarking with recent ~ed PV data at JLab[44] to reduce uncertainties. Their
result corresponds to a contribution to the asymmetry at Qweak kinematics that is
equivalent to a shift in the proton’s weak charge of 7.8 ± 0.5% of the tree-level SM
value.
30
CHAPTER 3
Experimental Setup
3.1 Experimental Setup
The experimental requirements associated with the Qweak experiment brought
about a unique set of technical challenges, due primarily to the proposed 2.1% sta-
tistical and 1.4% systematic uncertainty on the measurement of the part-per-billion
level asymmetry. In an effort to meet these goals, the experiment built on technical
advances made at Jefferson Lab, as well as expertise gained from previous precision
PVES experiments [45, 46]. The design parameters (shown in Table 3.1) were chosen
to optimize counting statistics and reduce systematics as much as possible. The pri-
mary limiting factor for the experiment was counting statistics, for this reason high
current, highly polarized beam, and a high power target were required, as well as
almost 2 years of non-continuous running. The major subsystems in the experiment
include: a 2.5 kW cryotarget, lead collimation system, large acceptance Cerenkov
detector array, toroidal spectrometer, rapid reversal polarized source, and tracking
system used to measure the experimental Q2. The experiment was run in two com-
31
Parameter Value
Incident Beam Energy 1.165 GeVBeam Polarization 85%Beam Current 180 µATarget Thickness 35 cm (0.04 X)Full Current Production Run Time 2544 hoursNominal Scattering Angle 7.9
TABLE 3.1: Design Parameters for the Qweak Experiment
plementary configurations: current-mode and tracking-mode. Current-mode, which
is the default running for the asymmetry measurement, relies on high current (180
µA) beam scattering from the lH2 cryotarget. Electrons scattered at a nominal an-
gle of 7.9 are focused into the radially symmetric detector array, using the toroidal
spectrometer, where they are integrated at 960 Hz and read-out via 18-bit ADCs.
In tracking-mode, which is run at low current (∼1 nA), horizontal and vertical drift
chambers are used to reconstruct the scattered electron track and measure the ex-
perimental Q2 on the detectors. A number of ancillary measurements were also
performed to quantify the contribution of background to the asymmetry measure-
ment. In the following chapter a more detailed overview of the implementation and
operation of the different subsystems is given.
32
3.2 Polarized Electron Beam
3.2.1 Continuous Electron Beam Accelerator Facility
The Continuous Electron Beam Accelerator Facility (CEBAF) provides high
quality polarized beam that is crucial to the PVES program at Jefferson Laboratory.
The design of CEBAF provides an electron beam that, while having some time
structure, is essentially continuous providing a duty factor of 100% [47]. CEBAF
consists of 2 parallel linacs comprised of 20 cryomodules which can produce 600 MeV
of acceleration using niobium Superconducting Radio Frequency (SRF) cavities. The
use of SRF technology greatly reduces the power consumption by reducing ohmic
heating that would be present otherwise. Each linac is attached to a total of 9
recirculating arcs; for each pass, electrons are selected using an RF separator into
an arc with progressively stronger dipole magnets to accommodate the increased
electron momentum while providing the same radius of curvature. At the exit of
each arc an RF combiner recombines the beam and sends it back into the linac for
acceleration. The beam is extracted using an RF separator, following the desired
number of passes, at the exit of the south linac. The beam then enters the beam
switch yard where is it sent to any of the 3 experimental halls. The race track design
of CEBAF, shown in Figure 3.1, provides a total of 5 passes with a final maximum
energy of 6 GeV which can be delivered simultaneously to all three halls with a
current in excess of 200 µA.
33
FIG. 3.1: CEBAF schematic.
3.2.2 Polarized Source and Injector
The production of spin-polarized electrons at Jefferson Laboratory begins in
the injector where circularly polarized light incident on a photocathode of gallium
arsenide (GaAs) is used to excite electrons from the conduction band to the vacuum.
The electron affinity of GaAs, χ, defined to be the energy difference between the vac-
uum level and the conduction band, is normally positive, however by treating GaAs
with a layer of Cs2O, χ can be reduced below zero giving a negative electron affinity
(NEA)[3]. While a NEA photo-cathode is not a requirement of a spin-polarized
electron source it significantly increases the emitted electron intensity and provides
substantially better quantum efficiency (QE). The polarized electrons are produced
using the process of optical pumping; when a photon with angular momentum σ+(−)
is incident on a GaAs photo-cathode electrons transition from the conduction band
substate P3/2 to the valence band substate S1/2. The transitions between mj=±3/2
and mj=±1/2 give a two-fold degeneracy which is illustrated in Figure 3.2. The
fractional transition probability between the P− 32,− 1
2→ S− 1
2, 12
and P 12,− 1
2→ S 1
2, 12
34
shows photo-electrons are 3 times more likely to undergo the first transition than
the second which limits the maximum theoretical polarization possible to 50%.
FIG. 3.2: The allowed optical transitions of ∆mj = ±1 in a GaAs photocathode areshown. The numbers in the circles represent the relative transition probabilities[3].
In order to raise the achievable polarization even further to accommodate ultra-
precise parity experiments, it is desirable to break the degeneracy of the P substate
allowing a theoretical maximum polarization of 100%. By growing the GaAs photo-
cathode on a single layer of GaAsP it is possible to apply a mechanical strain to
the photo-cathode breaking the degeneracy and pushing the polarization to ' 75%.
Going even further, and alternating thin layers of GaAs/GaAsP, a super-lattice
can be formed which can generate beam with a polarization of up to 90% and a
significantly improved QE [48]. With this, the injector at Jefferson Laboratory
routinely provides 85% polarized electron beam with a QE of near 1%.
The circularly polarized light used to produce the electron beam in the injector
is a product of the device chain shown in Figure 3.3. Three diode lasers are pulsed
at 499 MHz each, 120 out of phase, with a combined bunch frequency of 1497
MHz. The resulting light is linearized using a linear polarizer, then passed through
a Pockels cell which has a birefringence proportional to the electric field applied to
the crystal medium. This birefringence rotates the polarization vector component
35
differently depending on the axis, and therefore with the correct electric field the
light becomes circularly polarized. By rapidly reversing the applied voltage it is
possible to flip the polarization vector and thus reverse the direction of the electron
polarization from the photo-cathode. Imperfections in the birefringent crystal can
lead to residual linear polarization, which has the effect of adding a small amount
of ellipticity to the polarized light. Most optical systems transport one linear po-
larization more effectively than others resulting in a helicity-dependent asymmetry
in the amount of light delivered to the photo-cathode. This effect is traditionally
referred to as the polarization-induced transport asymmetry (PITA)[49]. This ef-
fect is magnified by the fact that the GaAs photo-cathode has an analyzing power,
i.e. emitted current from the photo-cathode is dependent on the orientation of the
major-axis of the polarized light. This helicity-dependent effect generates a certain
class of helicity-correlated beam asymmetries which include charge asymmetry.
FIG. 3.3: Schematic of CEBAF injector optics setup[4].
The Rotatable Half-Wave Plate (RHWP) is used to minimize helicity-correlated
asymmetries by rotating the residual linear polarization vector. This is done by
matching up the major-axis of the exiting laser light to the axis of the analysing
power such that the residual linear polarization is equal in magnitude in both polar-
ization states. The Insertable Half-Wave Plate(IHWP) flips the sign of the linearly
polarized light striking the Pockels Cell and therefore changes the sign of the circular
polarization leaving the Pockels Cell. This is an important slow reversal because
36
it cancels any systematics such as beam steering and electronic cross-talk which do
not change sign with the IHWP[4].
Polarized electrons extracted from the photo-cathode are accelerated into the
injector via the 100 kV DC gun, which is enclosed in an Ultra High Vacuum (10−11-
10−12 Torr). This reduces the effects of back-scattered ions in the enclosure colliding
with the photo-cathode which would degrade the QE. With this, the photo-cathode
can provide an integrated charge of approximately 500 C before there is a need to
restore the photo-cathode yield. Electrons accelerated into the injector are longi-
tudinally polarized parallel(antiparallel) to the beam momentum, however during
transport through the linacs and arcs the electrons exhibit an in-plane precession
which may result in a net transverse polarization at the experimental hall. This
effect can be removed using a Wien Filter which rotates the electron spin in the
injector such that it cancels the net spin procession in the accelerator[50, 51]. A
Wien Filter consists of an electric and magnetic field perpendicular to one another
and perpendicular to the beam momentum. Originally used as a velocity selector,
charged particles with a velocity satisfying β = |E/B| are passed without deflection
but with their spin vector rotated. Building on this, the Two Wien Filter Spin
Flipper at CEBAF shown in Figure 3.4 uses a Vertical Wien Filter to rotate the
spin 90 to the vertical, a pair of solenoids to rotate transversely into the horizontal
plane, and a Horizontal Wien Filter to completely reverse the initial electron spin.
This forms the basis for an entirely new method of slow reversal which allows the
electron spin to be flipped independently of the laser.
37
FIG. 3.4: The Wien filter causes a spin precession of the electron spin. The doubleWien filter uses a pair of Wiens, Vertical and Horizontal, to orient the electron spinso as to cancel out transverse polarization due to spin procession in the linac andarcs. This allows 100% longitudinal polarization to be delivered to the experimentalhalls[5].
3.2.3 Beam Position Monitors
Beam transport information in the injector and the Hall C line is provided by
beam position monitors (BPM), which play an important role in providing quality
electron beam at Jefferson laboratory. BPM’s are cylindrical cavities containing 4
symmetrical stripline wires rotated to 45 in the right-handed Hall C coordinate
system, which couple resonantly (1497MHz) to the RF-signal of the passing beam,
providing signal amplitudes proportional to the proximity of the beam to the signal
wires. The rotation of the signal wires is needed to alleviate the effects of synchrotron
radiation in the horizontal plane. The BPM readout electronics can be adjusted to
handle a range of currents(1µA−200µA) by adjusting the gain settings on the pick-
up wires. Each of these signals is transmitted to the BPM readout electronics where
it is amplified and down-converted to 1 MHz before being digitized and read out via
custom Qweak ADC’s at 960Hz in integrating mode. Figure 3.5 shows a cross-section
of a BPM cavity and stripline wires.
38
FIG. 3.5: Schematic of cylindrical BPM as seen along the beam pipe[6]. Signal wiresare shown rotated in the counter-clockwise direction to 45. Hall coordinates aregiven by XH and YH .
The relative beam position in the unrotated basis can be written in terms of the
raw signals in the form of the asymmetry between wires along a given rotated axis.
In order to properly calculate the absolute beam position two important factors
need to be addressed[52].
• The amplitude gains may be different between channels.
• The pedestals in each channel are not zero and must be accounted for.
In the unrotated basis the absolute beam position in X′ can be written in the form:
X ′ = κ(XP −XPoffset)− αX(XM −XMoffset)
(XM −XMoffset) + αX(XP −XP−offset)(3.1)
Here α is the ratio of the gains between two channels along a given axis, and κ is the
sensitivity of the BPM to the 1497 MHz RF-signal, which converts the readout signal
to mm. The Y′ position can be calculated in a similar manner. The pedestals are
defined as the BPM readout at zero current however, due to the non-linear behaviour
of the BPM’s at low current, they must be measured in a region of linear operation
and calculated using extrapolation. BPM calibrations are done at high current
(nominally 180 µA) in fixed gain mode by varying the beam current ±10−15%
39
and measuring the monitor response. A linear fit to the monitor response allows
extrapolation to zero current and robust determination of the pedestal value[6]. The
absolute beam positions read out by the data acquisition system are rotated into
the accelerator coordinate system,
X =1√2
(X ′ − Y ′ −X ′offset) (3.2)
Y =1√2
(X ′ + Y ′ − Y ′offset) (3.3)
where the offset variables shift the origin of the relative beam position into the
origin of the accelerator coordinate system.
The final BPM in the Hall C line is located 1.4 m upstream of the Qweak target
making direct measurement of the electron beam position at the target impossible.
In order to determine the beam position(angle) on target, virtual BPMs are con-
structed using the 5 BPMs in the drift region to project to the target. Virtual BPMs
are constructed using a linear least-squares fit to the absolute BPM positions. A
similar measurement of the relative energy at the target can be made by considering
the computed position(angle) at the target and the horizontal beam position at the
point of highest dispersion in the Hall C arc as measured by BPM 3C12X. The
functional form of the energy variable, qwk energy, is given by:
dPtargetPtarget
=X3C12X −M11Xtarget −M12X
′target
M15
(3.4)
where Mi,j are elements of the transport matrix which map position, angle, and
relative energy at the target to the corresponding measurements at the point of
highest dispersion in the Hall C arc. Here we have only considered elements of the
transport matrix which contribute significantly to the determination of qwk energy.
40
A list of the BPM’s used in the construction of the virtual BPM’s can be found in
Table 3.1.
Run Period BPMs used in Virtual BPM
Beginning of running to run 14486 3H07a, 3H07b, 3H07c, 3H09, 3H09b
Run 14486 to end of running 3H07a, 3H07b, 3H07c, 3H09
TABLE 3.2: Hall C BPMs used to construct virtual target BPMs. BPM 3H09b wasremoved due to functionality issues in the second half of running.
3.2.4 Beam Current Monitors
The electron beam current in CEBAF is monitored using beam current monitors
(BCM) in the injector and the end of the accelerator. Measuring the beam current
in two locations helps to monitor and limit beam loss through the machine, and
provide the most accurate measurement of the beam incident on the target. Current
measurements made on the accelerator side were done using a pair of BCM cavities in
conjunction with an Unser monitor. The BCM’s are cylindrical RF cavities resonant
at the fundamental frequency of the beam (1497 MHz). As the beam passes through
the BCM it excites the transverse electric mode, TE010 via a large loop antenna
coaxial with the cavity resulting in a DC voltage level linearly proportional to the
beam current. The BCMs are temperature controlled to reduce the effects of heating
due to power dissipation in the cavity. The high-frequency output signal is down-
converted from 1497 MHz to 1 MHz. The BCM signals are then read out into
the DAQ using Qweak ADCs. Table 3.2 shows the 6 BCM cavities active during
different run periods over the full experiment. For purposes of noise reduction,
a charge variable (qwk charge) was formed using the average of the most robust
current monitors during each run period.
41
Run Period Active BCM qwk charge
Run 1 BCM1, BCM2, BCM5, BCM6 BCM1, BCM2
Run 2 BCM5, BCM6, BCM7, BCM8 BCM7, BCM8
TABLE 3.3: Beam Current Monitors activity periods. The monitors used to formthe qwk charge variable are listed. BCMs 1−2 were used primarily during Run Iwhile BCMs 5−8 were being commissioned. BCMs 5−8 were built with lower noisedigital receivers.
The BCM cavities provide a relative current measurement and therefore must
be properly calibrated in order to make an accurate measurement. This done using
an Unser monitor. The Unser monitor is a parametric DC current transformer which
provides a non-invasive beam current measurement[53]. The Unser monitor contains
a small toroid sensor which develops a magnetic field as the beam passes through
it. The DC component of the magnetic field is detected and cascaded through a
feedback loop which generates a current in the secondary windings exactly canceling
the magnetic flux of primary windings. The current seen in the feedback loop is then
proportional to the detected beam current.
3.3 Electron Beam Polarimetry
Precise measurement of the longitudinal beam polarization is crucial to the
measurement of Qpweak. The uncertainty in the beam polarization measurement
represents the dominant relative systematic uncertainty contribution expected to the
physics asymmetry (δP/P = 1%). The beam polarization was measured using two
devices: The existing Hall C Møller polarimeter which provides an absolute invasive
polarization measurement at beam currents of ∼1µA, and the newly constructed
42
Hall C Compton polarimeter which provides a high current, non-invasive, continuous
measurement of the beam polarization.
3.3.1 Møller Polarimeter
The Møller polarimeter relies on cross-section asymmetry e−e− → e−e− scat-
tering for precise determination of the beam polarization. Møller scattering is a pure
Quantum Electrodynamic (QED) process so the cross-section can be calculated very
precisely. In the center-of-mass (CM) frame the scattering cross-section for Møller
scattering at lowest order can be written as [54]:
dσ
dΩ=dσodΩ
[1 + P||t P||b A(θ)] (3.5)
where dσo/dΩ is the unpolarized cross-section, P||t is the target polarization, P
||b
is the beam polarization, and A(θ) is the analyzing power. Considering the cross-
section asymmetry,
dσ↑↑o /dΩ− dσ↑↓o /dΩ
dσ↑↑o /dΩ + dσ↑↓o /dΩ= A(θ)P
||t P||b , (3.6)
the beam polarization can be obtained with knowledge of the analyzing power and
polarization of the target. The analyzing power is determined using simulation and
is maximized by detecting scattered and recoiled Møller electrons at 90. The Hall C
Møller polarimeter, shown in Figure 3.6, scatters longitudinally-polarized electrons
from a 1 µm thick iron foil target. The iron foil target is polarized to saturation
along the beam axis using a 3.5 T superconducting solenoid field. At saturation,
the electron polarization is known precisely[55]. Scattered electrons are horizontally
focused using quadrupole magnet Q1 and the desired scattering angles are selected
using the collimator system, after which they are again horizontally defocused using
43
quadrupole Q2. The scattered electrons are then detected in coincidence using two
symmetrically-placed lead-glass detectors. Detecting the scattered electrons in coin-
cidence greatly reduces contamination from Mott scattering and other backgrounds.
This allows measurement of the electron beam polarization with a statistical error
FIG. 3.6: Schematic for Hall C Møller Polarimeter[7].
of ∼0.5% in 5 minutes. Because the Møller measurement requires the insertion of a
special target into the beam path, and is therefore invasive to normal Qweak running,
Møller measurements were performed only periodically.
3.3.2 Compton Polarimeter
The Hall C Compton polarimeter provides continuous, non-invasive measure-
ment of the electron beam via e−γ → e−γ scattering. The detection of both the
back-scattered electrons and the photons provides two important, semi-independent
measurements of the electron beam polarization. The cross-section asymmetry for
Compton scattering is given by,
A(k) = PγPeAz(k), (3.7)
44
where A(k) is the cross-section asymmetry, Pγ is the circular polarization of the
laser, Az(k) is the analyzing power, and Pe is the electron beam polarization. The
Compton polarimeter, which was designed and commissioned primarily for Qweak,
consists of a magnetic chicane, laser system, a photon and electron detector, and a
laser monitoring system. The layout of the Hall C Compton polarimeter can be seen
in Figure 3.7. The incoming electron beam is directed into the Compton chicane
FIG. 3.7: The Hall C Compton layout. D1−4 are dipole magnets, the electron beamis shown in red, black denoted the scattered electrons, and the dashed blue showsthe scattered photons.
using dipoles D1−2 where it interacts with a photon target provided by a low-gain
Fabry-Perot cavity coupled to a 10 Watt continuous-wave green laser (532 nm). The
scattered electron are then deflected into the electron detector while the scattered
photons, which are emitted in a cone around the scattered electrons, are detected
in the photon detector which is located along the chicane axis. Unscattered beam is
directed out of the Compton chicane and back into the main beamline using dipoles
D3−4. The array of 4 dipole magnets in the Compton chicane are aligned in a
symmetric arrangement and have the benefit of deflecting the beam with no net
spin procession. Due to the Compton polarimeter being in commissioning it was
not available until April 2011, and is not considered in this analysis.
45
3.4 Primary lH2 Target
The primary scattering target for Qweak was a 34.4 cm long conical liquid hy-
drogen cell designed to sustain 180 µA of beam current, while keeping noise due
to target boiling <5% of the width if the physics asymmetry distribution. This
corresponds to a target noise contribution of <50 ppm based on the 5.3 GHz per
detector scattering rate in the experiment. Target heating, due to the incident elec-
tron beam, causes density fluctuations in the target, which can induce noise into
the measured main detector asymmetry. In order to reduce systematic effects due
to target heating, which can be a significant source of noise, the electron beam was
rastered in a 3.5 x 3.5 mm2 uniform distribution. The target length was chosen
to yield more scattering events, while the conical shape accommodates the optimal
scattering angle of 7.9. The lH2 target was operated at a pressure of 35 psi and a
temperature of 20.00 ± 0.02 K. The liquid volume of the target was approximately
55 liters when fully condensed. The target cell in which the incoming electron beam
interacts can be seen in Figure 3.8a and includes an entrance and exit window com-
posed of Al 7075-T6 which was chosen for its especially high tensile strength. The
target windows, which generate a significant amount of background, were made as
thin as possible. The entrance window has a uniform thickness of 0.096 mm, while
the exit window consists of two regions: an outer diameter which has a thickness
0.635 mm and an inner region of diameter 15 mm and a thickness of 0.125 mm. The
unscattered beam passes through the inner region, which is thinner to reduce back-
grounds. Elastically scattered electrons pass through the experimental acceptance
(6 < θ < 12) and exit the target through the thicker outer diameter of the target
window. The focus of scattered electrons, passing through the outer region of the
exit window, is affected due to radiative losses. Characterization and measurement
46
(a) Target cell drawing.(b) CFD simulation of target flowvelocities.
FIG. 3.8: (a) CAD drawing of the conical target cell. The conical shape of the cellaccommodates the out going scattered electrons. (b)CFD simulation showing lH2
flow velocities inside the target cell[8].
of these radiative losses was the subject of target window studies[8]. One of the pri-
mary concerns of high current beam on a lH2 target is heating; 1.165 GeV electrons
passing through the target experience an ionizing energy loss of ∼2.1 kW, which is
deposited as heat in the target. Additional contributions include: viscous heating,
pump heat, as well as conductive and radiative heat loss. Including 150 W of reserve
for the feedback loop, the total cooling power needed by the target is 2.5 kW [56].
To meet the cooling requirements of the Qweak target a large amount of effort was
put into the unique design of the target cell using Computational Fluid Dynamics
(CFD) simulations. The simulations effort help to optimize the circulation of the
lH2 in the target cell minimizing heating on the target windows and reduce hot spots
that would otherwise lead to target boiling. Examples of target flow velocity pro-
files from CFD can be seen in Figure 3.8b. The lH2 in the target cell is recirculated
at 1.2 kg/s transverse flow, using a 746 W centrifugal pump. A 3kW hybrid heat
exchanger, designed by the Jefferson Lab cryo group, was employed to help meet
target cooling demands. The heat exchanger combines two source of cooling, a 4K
supply from the Central Helium Liquefier, and a 15K supply from the End Station
47
Refrigerator, into a single unit. In addition to this, a high-power heater is used to
stabilize the target temperature when the beam is not present. The allowed contri-
bution to the systematic error of physics asymmetry, from target density fluctuation
due to target heating, was <50 ppm or ∼5% of the quartet level asymmetry width.
In order to measure the actual error contribution, a number of systematic studies
looking at dependence on raster size and lH2 pump frequency were performed. The
total contribution to the experimental asymmetry width due to target boiling was
found to be 46 ppm.
3.5 Infrastructure
3.5.1 Collimation System
A collimator system was used in the experiment to reduce the effects of inelas-
tics and secondary particles in the detectors, and optimize the uncertainty in the
measurement. The system consisted of three 15 cm thick lead collimators, each with
eight 400 cm2 apertures placed symmetrically around the beamline (see Figure 3.9).
The primary and tertiary collimators were designed to protect the tracking system
and the magnetic spectrometer from intense gamma radiation and inelastics coming
from the scattered beam. The secondary collimator (defining collimator) defines the
solid-angle acceptance for the experiment to be 4% of π in the azimuth and 49% of
2π in the polar angle. Although the primary function of the collimators is to define
the experimental acceptance and optimize the uncertainty of the measurement of
Qpweak, GEANT simulations found that the inner edge of the defining collimator -
which has direct line of sight to the detectors - was a potential source of background
in the form of high energy photons. This background, estimated to be on the order
48
of 1%, led to the design and installation of a set of lintels, which acted as an effective
fourth collimator, to block line of sight neutral backgrounds.
FIG. 3.9: The defining collimator shown before installation. This collimator definesthe experimental acceptance for Qweak[8].
3.5.2 Qweak Toroidal Spectrometer
Particles within the specified acceptance exit the collimator and enter the Qweak
toroidal (QTOR) spectrometer where charged particles are focused azimuthally to-
wards the detectors. QTOR is specifically optimized to focus elastically scattered
electrons, at the Qweak beam energy and angle, onto the detectors; inelastics and
Møller electrons, which have a lower energy are swept away. Positively charged par-
ticles are bent inward and absorbed by the thick shielding surrounding the beamline.
The QTOR spectrometer (Figure 3.10a) provides a toroidal magnetic field of ∼ 0.9
T-m and consists of eight water cooled coils, situated symmetrically around the
beamline, held in place by an aluminium support structure making it completely
iron free. The initial field mapping of the magnet was done at MIT-Bates after
which the magnet was disassembled, transported to JLab, and reassembled. After
installation in the experimental hall at JLab the magnet was remapped and cali-
brated to ensure precision installation. The reassembly was found to be within ± 5
mm of that found at MIT-Bates [57].
49
(a) QTOR during installation. (b) QTOR Diagram.
FIG. 3.10: (a)Qweak Toroidal Spectrometer installed in experimental Hall C at Jef-ferson Lab. (b) Current in each coil travels in a racetrack fashion creating a toroidalmagnetic field. The field goes to zero at the center allowing the beam to passunperturbed to the beam dump.
3.5.3 Shielding Wall
An 80 cm thick concrete shielding wall (Figure 3.11) was placed downstream of
QTOR, preceding the detector array. The goal of the shielding wall was to provide an
aperture which would reduce backgrounds resulting from secondaries and inelastics
interacting with structural supports in QTOR. GEANT 3 Simulations showed that
in the region of the shielding wall there was an overlap between the elastic and
inelastic signal; great care was taken in the design of the shielding wall apertures to
assure that, while blocking backgrounds from the spectrometer region, backgrounds
resulting from interactions of the inelastic signal with the inner edge of the aperture
were as small as possible[9].
3.5.4 Quartz Cerenkov Detectors
The main detectors for the experiment were responsible for measuring the rate
of elastically scattered electrons in each helicity state, which is used to measure the
PV asymmetry. Each main detector had dimensions of 200 cm x 18 cm x 1.25 cm
50
FIG. 3.11: CAD diagram of shielding wall. Simulated beam envelope shown in blue.Proper design of the shielding wall was crucial to avoid interaction of the scatteredenvelope with the inner edge of the apertures[9].
and was composed of Spectrosil 2000 synthetic fused-silica bars. Each detector bar
was a composite of a pair of 100 cm pieces of fused silica glued together with UV-
transparent glue (SES-406). The Spectrosil 2000 synthetic fused-silca was chosen
for the detector material on the merits of its radiation hardness, low sensitivity to
neutral background, and uniform response across the detector. Each detector bar
was placed in a light tight box and supported by an aluminium support structure.
The “ferris wheel”support structure held one detector package at a radius of 335
cm in each of the 8 octants (0, ±45, ±90, ±135, 180). A diagram showing the
azimuthal placement of the main detector packages is shown in Fig. 3.12. Scattered
electrons leave the collimator apertures and travel into the main detectors; elec-
trons moving faster than the phase velocity of light in the quartz medium generate
Cerenkov radiation. The Cerenkov light is internally reflected down the bar, through
a light guide, and into photo-multiplier tubes (PMT) mounted on each end. Each
light guide was a simple extension of the main quartz bar with the PMTs attached
to the ends using optical glue. The 5” PMTs gathered light using UV-transparent
glass and operated in two modes depending on the PMT base used: low-gain mode
(gain O(103)) which was used for high current integrating mode, and high-gain (gain
O(106)) mode which was used primarily for tracking mode. Signals produced in the
51
FIG. 3.12: Schematic of quartz detector array.
PMTs were read out using low-noise electronics designed and constructed at TRI-
UMF. A current-to-voltage preamp converts the current produced in the PMT to
an analog pulse. It is then integrated and recorded via an 18-bit ADC operating at
500 Hz. Bench tests of the low-noise electronics showed a noise contribution to the
asymmetry width of 7 ppm.
One of the primary concerns mentioned in previous sections has been the con-
tribution of soft backgrounds to our detector signal. Simulation indicated that sig-
nificant backgrounds would be present during high current running - photon rates
due to bremsstrahlung were predicted to be O(200MHz)[58]. Photons of energy less
than O(100keV) should not produce a signal in the detectors[59], however at higher
energies Compton scattering and pair production can occur inside the quartz and
electron-positron pairs can produce a background signal. In order to reduce the
effect of these backgrounds in the quartz detectors, 2 cm thick lead pre-radiators
were added to the main detectors. This has the effect of reducing backgrounds by
absorption of soft photons and increasing the signal-to-noise ratio from electrons in
52
the elastic envelope showering inside the pre-radiator. The pre-radiator did however
induce a level of noise in the detectors (∼12%) due to light-yield fluctuations in the
number of electrons produced in the shower; this increased the amount of run time
need by 370 hours but was out weighed by the reduction in the relative background
versus not using the pre-radiators.
3.5.5 Beamline Shielding
Scattered electrons leaving the scattering chamber at small angles (0.75 − 4)
and passing through the first clean-up collimator can potentially interact with the
downstream beam pipe causing a significant source of background - primarily in the
form of neutrals[8]. While efforts were made to design the Qweak beamline to ac-
commodate this, simulations showed significant background generation. A tungsten
beam plug was installed in the downstream side of the first collimator to block these
events. The plug was water-cooled to dissipate ∼1.3 kW of expected heating due to
the scattered beam being dumped into the plug. Figure 3.13 shows a simulation of
the beamline background with and without the tungsten plug; it is apparent from
the simulation results that the addition of the tungsten plug significantly reduced
both the neutral and multiple-scattered electron contributions to the beamline back-
ground. The tungsten plug, however, becomes a source of secondary photons in the
collimator region with potential to re-scatter and generate background in the detec-
tors. Measurements to determine how to best reduce backgrounds coming from the
beamline were performed during the commissioning phase, and led to the installa-
tion of a number of lead bricks around the beamline as well as a 12” long, 2” thick
lead donut around the beam pipe in the collimator region.
53
FIG. 3.13: Simulated beamline background with(right) and without(left) the tung-sten plug. Electrons are shown in red and neutral are shown in blue. The additionof the tungsten plug drastically reduces the beamline backgrounds[8].
3.5.6 Beam Modulation System
Changes in the beam properties during helicity flip can lead to false asymme-
tries in our detected scattering rates. The scattering rate in the detectors depends
significantly on five parameters: transverse position, angle, and incident energy E.
Small helicity-correlated variations in these parameters produce false asymmetries
which are potentially enhanced by various broken symmetries in the experimental
apparatus. The experiment took a two-pronged approach to measurement of detec-
tor sensitivity to changes in the beam position, angle, and energy. The first method
was to measure the correlation of the measured asymmetry in the detectors with the
helicity-correlated difference of the natural beam jitter seen in the beam monitors
and to use standard multivariate linear regression techniques to extract the decou-
pled detector sensitivities. The second - and a major topic of the work found in this
thesis - was to modulate the beam parameters in a controlled way using an external
driving signal and extract the sensitivities from the observed detector response. In
order to achieve this, a beam modulation system was designed and installed for the
Qweak experiment. The details of the design, implementation, and analysis efforts
using this method are detailed in the following chapter.
54
3.6 Tracking System
The parity-violating asymmetry, at Qweak kinematics, can be expressed in pow-
ers of Q2 (see eq. 2.25); therefore, precise determination of Q2 is essential to the
extraction of the weak charge. Specifically, the acceptance-weighted distribution of
Q2, weighted by the detector response needs to be determined to ∼1%. This is
important because the non-uniformity of light collection in the detector affects the
measured Q2 at the percent level. The goal of the tracking system was to mea-
sure the acceptance-weighted Q2, and use it for a benchmark simulations which can
be used to calculate Q2 at the scattering vertex. As a secondary motivation, the
tracking system can be used to study various backgrounds including scattering from
the target windows that may contribute to the measured asymmetry. The tracking
FIG. 3.14: Simple schematic of the experiment showing the 3 tracking regions.Simulated trajectories of elastically scattered electrons are shown in red
system consists of two regions: Region II, and III and was operated at low currents
(50 pA - 50nA) due to limitations on the rates that the detectors in Regions II and
III could handle. The low rates allowed the detectors to be run in counting-mode,
meaning that individual detected particle could be recorded and used in track re-
construction. Region II contained a set of horizontal drift chambers (HDCs), which
55
were used to determine the scattering angle and initial particle trajectory. Region
III consisted of a pair of vertical drift chambers (VDCs), which provided particle
position and angle information after the magnetic field. Attached to each pair of
VDCs was a trigger scintillator, which provided the tracking event trigger. Detector
packages in Region II and III were mounted in opposing octants on rotatable support
structures which allowed for a complete mapping of the main detector acceptance.
Also, the rotating support structures allowed the tracking detectors to be moved
radially out of the beam during high current running. A simple schematic of the
three tracking regions is shown in Figure 3.14.
3.6.1 Horizontal Drift Chambers
The Region II HDCs are located just upstream of the last collimator and provide
position and angle information about the scattered electrons entering the magnetic
spectrometer. The HDCs were designed to have a position resolution of ∼200 µm
and an angular resolution of ∼0.6 mrad. The active area of the HDCs was 38 cm ×
28 cm and each HDC unit contained 6 wire planes (u, v, x, u′, v′, x′). Each wire plane
contained 192 sense wires; wires in planes (x, x′) were strung horizontally, wires in
planes (u, u′) were angled by 53.2, and wires in planes (v, v′) were angled by -53.3.
Each wire plane was situated between two aluminized Mylar plates held at -2150
V, while the sense wires were grounded. Each chamber was sealed air-tight and
filled with a 65%- 35% argon- ethane mixture; the argon/ethane mixture is stable
and is easily ionized. The gas mixture was bubbled through isopropyl alcohol to
reduced the effects of chamber ageing. A total of 5 chambers were built with the
intent of using 4 chambers in the experiment with the 5th as a back-up. Each HDC
“package” consisted of two chambers mounted with a 40 cm offset along the beam
56
FIG. 3.15: Region II HDCs installed downstream of QTOR in Hall C.
axis. The HDC packages were mounted in opposite octants of a human-powered
rotating support structure, the region II “rotator”, which allowed the HDCs to be
used to map the full acceptance. The region II rotator also allowed the HDCs to
be retracted radially inward - out of the scattered electron envelope - during high-
current running. Charged particles passing through the HDCs ionize the gas; ionized
electrons drift towards the sense wires. Drift electrons nearing the sense wire begin
an avalanche increasing the total charge - and therefore the signal size - seen when
the wires are read out. A photo of the region II HDCs installed in Hall C can be
seen in Fig. 3.15.
3.6.2 Vertical Drift Chambers
The VDC’s, located downstream of the QTOR and the shielding wall, served
to provide information about the electron’s position and angle after leaving the
magnetic field of QTOR. Each drift chamber had an active area of 0.914 m x 2.438
m and contained two wire planes, labeled u and v, each with 279 wires. The wires
in each plane were strung at an angle of 26.45, which was found to be the optimal
57
angle using the simulation packge Garfield[10, 60], and with a spacing of 11.12 mm.
A full detector package consisted of 2 wire planes, 3 Mylar foils, 2 gas planes, and
one spacer frame (see Fig. 3.16b). Each frame was made from G10-FR4, which is a
sandwich of glass fibers and epoxy chosen for its ability to withstand compression,
strong dielectric properties, and minimal gas absorption. These properties were
important because each chamber was filled with a 50%- 50% argon- ethane gas
mixture and had to be air-tight. The Mylar cathode planes were held at -3800 V
and each sense wire in planes u and v was held at ground.
(a) Vertical Drift Chamber
(b) Layering for one VDC. The HVfoils are shown in magenta and wireplanes are shown in green. The cen-ter foil was aluminized foil on bothsides to provide field in both cells ofthe VDC.
FIG. 3.16: (a) Shows one completed vertical drift chamber in the lab at College ofWilliam and Mary. The VDCs were layered (b) and contained 2 wire planes, 3 HVplanes, a spacer frame, and 2 gas frames[10].
Each VDC pair was mounted on a large rotatable support structure, the Region
III “rotator” (the topic of the next section), in opposing octants. As with the HDCs,
the VDC pairs could be rotated to all octant pairs, allowing for full coverage of
the azimuthal acceptance, as well as being retractable, which allowed them to be
moved out of the scattered envelope during high current running. The chambers are
58
mounted at a 24.4 angle with respect to the support structure mounts and therefore
have a smaller angular acceptance than the HDCs but a higher spatial resolution
(< 200µm).
Signals in the VDC are produced in much the same way as with the HDCs;
scattered electrons entering the VDC at an angle and ionize the gas in each VDC’s
“cell”. The free electrons in the gas are then accelerated along the electric field lines
towards the ground wires; as the free electrons drift they ionize more of the gas,
creating more free electrons, cumulating in an avalanche when reaching the wires.
For a single VDC package there are a total of 558 individual channels to read out
or 2232 channels for the entire Region III tracking system. The read-out of such
a large number of channels posed a significant logistical and cost challenge for the
experiment. In response to this issue the detector signals were multiplexed, which
reduced the number of channels required by a factor of 9.
FIG. 3.17: Cross-section of one VDC “cell”. HV planes are shown in green and theparticle track is shown in blue[10]. Particles entering the cell ionize the gas causingan avalanche of charge moving towards the signal wires. The timing of these signalsis used to extract track information.
59
(a) Diagram of LVDS-ECL signalmultiplexing in the MUX crates.
(b) Time difference data from benchtesting.
FIG. 3.18: (a)Diagram of LVDS-ECL read-out and signal multiplexing in the MUXcrates. A single delay line is shown for the first four multiplexed signals. (b)Typicaltime difference data from bench testing of the MUX crates. Each peak represents awire in the delay chain; peak separation is ∼1.3 ns with an average σ of 80 ps[10].
Each wire signal was read out using a custom pre-amp/discriminator board
which amplified the analog signal and converted it to a low-voltage differential logic
signal (LVDS). The LVDS signal is then converted to an emitter-coupled logic (ECL)
signal and fed into a custom JLab 64-channel time-to-digital converter (TDC) which
recorded the arrival time of the incoming logic signal. The signal multiplexing took
place in custom built multiplexing (MUX) crates containing LVDS-ECL conversion
boards. The multiplexing was done by combining the signal from every 9th wire into
a single read-out channel with a small delay provided by the ECL gate/buffer chips.
An example of the signal read-out can be seen in Fig. 3.18a. Each signal was split
and processed by the LVDS-ECL chip; signals were then read out on both the “left”
and “right” sides.
Depending where along the signal chain the signal is it will encounter a different
number of delay buffers (each delay is ∼1.3 ns). The signals were processed by
subtracting the times for the left and right signals, which produced a well defined
peak for each wire, as seen in Fig. 3.18b.
60
3.6.3 Trigger Scintillators
The trigger scintillators served to provide a fast timing trigger for tracking-
mode data and were located 40 cm upstream of the main detectors. Each trigger
scintillator was held in place using aluminum supports attached to the Region III
VDCs. Each scintillator bar was 218.45 cm x 30.48 cm x 1.00 cm thick and was
made from Bicron BC-408 plastic scintillator which is sensitive to charged particles
and insensitive to neutrals. The scintillator bars had light guides made from UVT
lucite stranded and attached to each end; the UVT lucite strand was choosen over
a triangular light guide due to a factor of 2 increase in their light collection and
30% increase in timing resolution[8]. On the end of each light guide was attached a
FIG. 3.19: Schematic of trigger scintillator including dimensions[8].
Photonis XP-4312B 3” high gain (∼ 3x107) PMT which converted the scintillator
light into an electrical output signal. The signals produced by each PMT were
combined using a CAEN V706 16-Channel Mean-time Module, which produces an
61
output signal independent of the hit position along the length of the scintillator
bar. This eliminates the time difference between signals seen in PMTs at each end
of the trigger scintillators - and provides a constant timing signal - in the case a hit
is closer to one side than the other.
3.7 Rotator
As mentioned previously, both the Region II and Region III tracking detectors
had access to the full azimuthal acceptance through rotation of their support struc-
tures. One stark difference between Region II and Region III was that while the
Region II HDCs - each HDC weighed ∼25 kg - were easily rotated by hand, the
Region III VDC package (VDCs plus mounting plates) weighted ∼953 kg and were
rotated around a horizontal axis 3.96 m vertically from the hall floor and at a radius
of 2.3 m. In this case, manual rotation was not a viable option and a mechanically
driven rotation system was designed. The Region III rotator was designed to pro-
vide a semi-automated method of rotating the Region III VDCs through the full
azimuthal acceptance, to provide radial motion that can move and retract the VDCs
during high current running, and to provide sufficient reproducibility given the res-
olution of the VDCs. The system included radial and rotational motion systems,
detector cabling management, position tracking, an air-driven position locking sys-
tem, and a safety system. Each of these systems, their design criteria, and measures
of their performance will be discussed at length in the following subsections. More
detailed specifications will be listed in the appendix.
62
FIG. 3.20: Region III rotator.
3.7.1 Structural
The Region III rotator (see Fig. 3.20), at its center, consists of a 2.3 m radius
304 stainless steel hub which is held concentric to the beamline. The 304 stainless
steel, which makes up the structure of the rotator, was chosen for its low magnetic
permeability1. The central rotator hub is held in place via lateral struts attached to a
simple steel support base. Each lateral strut has two 5-inch V-groove rollers attached
to a base allowing them to be adjusted radially and in Z along the beam direction.
The inner edge of the central hub is beveled to fit the V-groove cams. During
installation the rotator hub was situated on the adjustable rollers, and a survey was
used to adjust the centroid of the rotator hub to be centered on the beamline. The
1The low magnetic permeability material was chosen to reduce effects due to polarization ofthe support structure caused by the near-by QTOR field and polarized electron scattering.
63
pitch was also adjusted to eliminate any tilt in rotator face with respect to the plane
perpendicular to the beam direction. Two parallel 2.88 m rails are mounted on the
FIG. 3.21: The central hub of the Region III rotator. The rotator support hubwas machined from 304 stainless steel; this material was chosen because of it lowmagnetic permeability. Steel rails, which the rotator arms ride on, were attached tothe flat structures protruding from the central hub.
central hub. The rotator arms, designed and assembled by Jefferson Lab, can be
seen in Fig. 3.22a. Each arm had two 35.56 cm threaded rods attached to the ends
on which the VDC packages were attached. The rotator arms were attached to the
central hub via four linear motion bearing blocks. These bearing blocks allowed
the arms to slide along the mounting rail during retraction(extension) of the VDC
packages. The arms were moved along the rails using a set of linear stepper motors
mounted to the slider supports. The linear motion system is discussed in more
detail in Sect. 3.7.2. The full detector assembly consisted of the two mated steel
‘Z’ plates to which the VDCs were bolted. The Z-plates held the VDCs offset from
each other in the Y-axis and at an angle with respect to the plane of the rotator
arms. The full detector assembly was mounted on the arms using the 35.56 cm long
threaded rods; the mounting bracket on the Z-plates, to which the threaded rods
64
were attached, were adjustable in (X,Y, Z ). This allowed the chamber’s positions to
be finely adjusted to the proper location with respect to the collimator apertures.
(a) The sliding arms can be seen in the HallC before being installed. The threaded rodswhich were used to hold the VDC packageswere covered with steel tubes prior to instal-lation. The mounting brackets for the pan-cake cylinders can be seen in the back right.
(b) CAD diagram of the fully as-sembled VDC packages mounted be-tween the slider arms.
3.7.2 Linear Motion System
The Region III detectors were only used during tracking measurements and cer-
tain focused ancillary measurements; the chambers were designed to be operational
at charged particle rates O(kHz) per octant and were retracted during high current
running which had rates O(GHz). The linear motion of the VDC detector packages
was achieved using a linear motion system which relied on two linear stepper motors
(EC35005B ) from IDC Motion. When choosing a stepper motor two things were
taken in to consideration: the linear motors needed to have the force required to
move the chambers and the position resolution in line with what was needed to de-
termine the track position in the VDCs. To make the linear motion safer and easier
the decision was made to only operate the linear motion system while the chambers
65
were in the horizontal position; this reduced the load on the motors to a minimum.
The required thrust load needed was estimated assuming a coefficient of friction
between the arms and the rails of 0.5; this corresponds to steel on steel. This is of
course an over estimation, but gave us sufficient overhead so that we would easily
be able to move the chambers. The estimated weight of the VDC detector package
(two VDCs plus support plates) was 9346 N which would give a required linear
thrust of 4673 N; the maximum thrust specification on the linear stepper motors
was 7200 N which is well above what was required. The position resolution on the
motors was rated to be 0.039 µm/step with a repeatability of 13 µm, which was well
below the ∼200 µm VDC position resolution. Each linear motor was driven using
a stand alone motor controller (S6961). The motors were controlled via a RS232
communication port on a single board computer located in the controls crate. The
motor controllers were wired in series so that the serial communication commands
would be sent to both controllers at the same time, thus ensuring the motors would
not become out of phase (see figure 3.23). While the rotator hub was locked in
place using a positioning pin during linear motion, an unbalanced load due to the
motors not being in sync was not desirable. The controls system will be discussed
in more detail in Sect. 3.7.4. The linear motors were mounted using a specially de-
signed motor mount to the central hub over the sliding rail to which the arms were
attached. The linear motor shafts were attached to the arms using stainless steel
eyelets welded to the posterior side of the sliding arms; each motor shaft had a clevis
mount at the end which could be attached to the eyelet on the arms. When the
rotator arms were extended they were held in place using 2.5 cm diameter locking
pins. Each pin - two per arm located on opposite sides - was driven into a pin block
mounted on support structure below the sliding rails. Each pin block was stainless
steel with a brass insert in the center. The brass insert was added due to problems
66
FIG. 3.23: Schematic showing the communications diagram for the linear motorcontrols.
with actuating the locking pins. When the locking pins were inserted and the arms
were rotated into a position, where a substantial amount of the VDC weight was
placed on them, pressure welding would take place making it extremely difficult to
remove them. Adding the brass insert to the pin block provided a buffer between
the steel pin and the pin block material.
3.7.3 Rotational Motion System
The rotational motor system provided ±180 rotational freedom to the VDCs;
this provided full azimuthal coverage of the detectors in each octant pair. The
rotational motor system is made up of a simple chain-drive powered by a 3 hp AC
motor. For the rotational motor an AC motor was chosen over a stepper motor
due to the need for a significant amount of starting torque. The inner face of the
central rotator hub was fitted with nine 2-row sprocketed face plates spaced in 40
increments around the rotator face. The AC motor was coupled to the sprocket
rotator face using a steel double-stranded roller chain. The initial design of the
rotator face only called for incremental face plates to be added to the rotator face,
67
however in the rotator construction phase there were significant issues with sagging
of the roller chain between segment. To address this issue spacer plates were added
to the rotator face to provide support between sprocket segments. Because the
roller chain length is given in increments of a single link, there was the possibility
that it would not be sufficiently tight and slipping might occur when the chambers
were being rotated. To address this problem the mount for the AC motor was
made to be adjustable in 3 independent degree of freedom. During installation the
motor was mounted and slack was removed from the chain; the motor was then
shimmed and bolted into place. One thing that was extremely important with
(a) Rotator face. (b) Sprocketed face plate.
FIG. 3.24: Sprocketed plates placed along the rotator face were used along with achain drive to turn the rotator.
the rotational motor was that rotation be done at low speed; included with the
rotation motor was a speed reducer that provided a reduction of 249:1. This gave a
maximum rotational frequency of 7.02 rpm. The original design criteria was that the
rotational frequency be not more than 0.1 - 0.2 rpm. Further reduction was achieved
by considering the gear ratio between the rotator hub and the size of the gear used
for the rotational motor. Given the rotator “gear” radius of 2.895 m, the motor gear
was chosen to be 0.075 m giving a gear ratio of 0.026:1. This reduced the rotational
68
frequency to a maximum of 0.183 rpm. The rotational motor was controlled using a
variable frequency drive (HF-320α Series). The drive could be controlled a number
of ways, including: manual control using a potentiometer on the front panel, remote
communication via RJ45 cable, or a 0 - 10 V terminal connection. The rotator
controls system made use of both manual and automated control using terminal
connections; details of the controls system set-up and implementation are discussed
in more detail in Sect. 3.7.4. In the same manner as the linear arms, the rotator
hub was locked in place after moving between octants. In each of the eight octants
there was a 1 in. diameter pin hole on the upstream side of the rotator face (an
example of this can be seen in Fig. 3.24a). In the same way as the linear arms, these
pin holes were used to lock the rotator into the desired octant using an air-powered
pancake cylinder. These cylinders drove a 1 in. diameter beveled steel pin into the
pin hole. Confirmation of the pin being in was handled using a standard door jam
switch - these are the same switches used in cars to detect whether a door is open.
In each of the eight octants, on the backside of the upstream rotator face, a door
jamb switch was mounted via a small hole drilled into the rear side of the rotary
locking pin holes. When the rotational pin was inserted into the pin hole of a given
octant, the door jamb switch was triggered and provided a +5 V signal which could
be read out by the controls system.
3.7.4 Motion Controls System
The motion controls system for the Region III was designed to provide a safe,
effective way of controlling the linear and rotational motion of the rotator. Both
the linear and rotational motor systems were manually operable, however the job of
the control system was not only to provide a conduit to control this motion, but to
69
provide synchronous linear motion, fail-safe controls, and more repeatability than is
easily achieved when triggering motion by hand. The rotator controls system (Figure
3.25) was centralized in a standard controls rack located inside the entrance to the
Region III shielding bunker. This location allowed operators to be in line-of-sight of
the rotator while being safe from any possible failures. The main components of the
controls system were the Versa Module Europa bus (VME) controls crate, Linear
Motor Drives (S6961), and variable frequency drive. The VME crate contained an
MVME6100 single board CPU, VME 612 Digital to Analog converter(DAC), a cus-
tom designed VME 16 channel relay board, custom 16 channel digital input register,
and an MVME761 Serial input-output board. The linear and rotational locking pins
FIG. 3.25: Motion Controls Rack.
were operated using air driven pancake cylinders. Operationally, these cylinders are
very simple; each cylinder has two valves, depending on the valve to which air was
70
applied a central pin was moved in or out. The valve not having air applied to it
became an exhaust valve. The air for these valves was routed from a high pres-
sure source in the Hall C and was regulated to 100 psi - this is the recommended
maximum pressure for these cylinders. The actuation of the pins was controlled by
voltage switched pneumatic valves mounted on the inner hub of the rotator. Control
of the linear motors was done using the rotator controls libraries running locally on
the MVME6100 CPU. The controls libraries were loaded automatically when the
controls crate was booted. Controls functionality was accessed by first using a secure
shell (SSH) login to the MVME6100; functions in the rotator controls libraries were
accessible from command prompt once an SSH session was initiated. Both software
and hardware checks were applied to linear motion commands to ensure that the
proper order of operations was observed during motion. Software monitoring was
done through the 16 channel digital input register. The function of this board was
to keep track of the voltage level on a given input channel, and therefore the state of
the sensor attached to it. The state of each input was then accessible by reading the
memory register associated with the channel. For the linear motion, the position of
the VDC was cataloged and queryable via serial communication to the motor drives.
Limits on the maximum and minimum position of the VDC were placed using Reed
switches; the Reed switches operate on the principle of magnetic proximity. Once
the motor drive shaft was moved to one of the limits, an internal magnet on the
shaft triggered the normally-open (NO) Reed switch, stopping the motor. In order
for linear motion to be possible three conditions were required: the rotator had to
be in the horizontal position, the rotational locking pin must be inserted, and the
linear locking pins had to be removed. The horizontal position was checked using
a tilt sensor which was normally-open in any position other than horizontal. The
checking of the rotational pin was done using the pin-in sensor located inside of each
71
rotational pin hole. The linear locking fail-safe was managed using both the input
register and the relay board. When the software command was given to engage
the linear locking pins the associated relays were closed on the board which applied
power to the pneumatic valves managing the air-driven pancake cylinders. Due to
the possibility of software failures, the linear motion system controls were tied to a
physical relay AND of the aforementioned conditions. The power to the linear motor
was wired to a relay in the NO position. Power to this relay was physically tied to
all three of the conditions above being met; in this way the linear motors were shut
off completely until it was safe to move them. Similar to the implementation of the
linear motion controls the rotation controls were given both software and hardware
fail-safes. The rotational motions controls code used read-back from the rotational
locking pin, linear locking pins, and linear motor brake to determine whether or not
it was safe to rotate. As an added level of precaution the rotational motor controls
were also tied to a physical relay fail-safe. The conditions for the rotational motion
to be operational were: power applied to the linear locking pins, rotational locking
pin in the out position, and the brake applied to the linear motor. The power to
the variable frequency drive that controlled the rotational motor, as well as pow-
ering it, were tied to a high voltage relay; power to this relay was dependent on
a physical AND of the relays controlling the preceding conditions. Originally the
rotational controls system were designed to be much more automated. When first
implemented, the controls system kept track of what octant the rotator was in, and
could be moved to other octants simply by giving the command and desired octant
number. This mode of operation was not used during the experiment, however, for
two reasons: 1) during installation the support bracket holding the tracking sensor
was mis-welded. and 2) the rotation of such a large structure, in a region containing
not only the VDC packages but the main detector for the experiment was deemed
72
much safer if done by hand. No longer using the sensor that provided a read-back of
the rotational position made moving to a given octant and securing the rotational
locking pin more difficult. With this in mind, an improvised positioning system was
constructed. A number of standard laser pointers were mounted on the main detec-
tor support structure (see Fig. 3.26). While the rotator arms were in the extended
position, survey points were marked on the VDC plates and rotator arms where the
laser spots were; by trying to align to multiple points in each octant we hoped to
decrease the error on the repeatability in each octant. The size of each laser spot
was ∼ 1 mm diameter which gave an upper limit on how well we could ensure the
rotational repeatability.
FIG. 3.26: Laser pointers mounted on main detector support structure.
3.7.5 Performance and Repeatability
During the 2 year run period of the experiment, the region III rotator performed
well, being one of the few subsystems with no major failures. Along with reliabil-
ity, the key metric for the performance of the rotator was its repeatability; poor
73
repeatability increases the error on how well we are able to determine the scattered
electron trajectory through the chambers. Both during commissioning and between
run periods repeatability studies where completed with the help of the survey group
at JLab. A listing of rotator surveys as well as notes about the program of each
can be found here[61]. The survey study presented in this thesis was completed in
September 2010 prior to the completion of the commissioning running. The results
of the study showed the radial consistency between octants, i.e. the difference in
the measured radius of each measurement point as the VDCs were rotated between
octants, to be sub-millimeter in the horizontal position and at the millimeter level
in the vertical position. The radial numbers were larger when only considering the
xy-plane (perpendicular to the beam direction) indicating the VDCs were tilted in
the direction of the beam. The angular repeatability when rotating the VDCs away
from the horizontal position and back was found to be on the millidegree level. The
repeatability of the VDC after being retracted and extended was found to be good
to the 100 µm level. Lastly, the distance between measurement points was stud-
ied. The survey is done by attaching tooling balls in special positions on the face
of the VDC packages, because these tooling balls are mounted in static locations
the distance between them should not change within the resolution of the survey
measurement and any contributions from potential sagging or flexing of the cham-
bers. The positional difference between points was found to be good to the micron
level. The full summary of results as well as a more in-depth discussion of each
measurement can be found in A.2. The results of the rotator study show that the
repeatability of the chambers position was well within the limits of what is required
of our Q2 measurement.
74
CHAPTER 4
Beam Modulation
4.1 Modulation System
4.1.1 Beam Modulation and Helicity-Correlated Beam Sys-
tematics
One of the more challenging aspects of PVES experiments is suppression and
measurement of helicity-correlated beam systematics (HCBS). These changes in the
beam intensity, position, profile, and energy - when correlated with the helicity of the
beam - create false asymmetries that distort the measurement of the physics asym-
metry. These false asymmetries are especially important in precision experiments
due to the small asymmetries that are being measured. One of the primary source
of HCBS is imperfections in the laser polarization; laser light on the photo-cathode
- while highly polarized - can contain small components of linear polarization which
switch sign under helicity flip. This, along with anisotropy in the quantum efficiency
of the photo-cathode, can cause helicity-correlated differences in the intensity of po-
75
larized electrons being produced from the photo-cathode. This shows up as a charge
asymmetry in our measurement. Other sources of HCBS include lensing, where the
Pockel cell operates as an electro-optical device deflecting the beam on helicity flip,
electronic noise, and helicity signal leakage in the injector[4]. A number of well
established methods were used to remove or heavily suppress HCBS; methods such
as the RHWP and the IHWP work to reduce HCBS sources both before and after
the Pockels cell as well as providing a diagnostic which helps determine their origin.
The central purpose of the following section is the determination of the detector
sensitivity to helicity-correlated beam motion using the beam modulation system.
The main benefit of using the beam modulation system was that it provided a
measurement that was directly correlated to the driven motion of the beam. The
alternative method, linear regression of natural beam motion, provides an effective
way to reduce the asymmetry width and remove correlation of the detector response
to beam jitter, however the meaning of these corrections is not well-posed. Take
for example a hidden variable, such as halo, that is correlated to the jitter in the
beam on helicity flip; this would give the extracted sensitivities a different character
detracting from our understanding of the HCBS. The result with modulation is much
clearer: removal of the correlation of detector signal to a controlled beam motion.
Hidden variables should not contribute, and any residual correlations would indicate
non-linearities, measurement failures, or hidden variables.
4.1.2 Beam Modulation System Instrumentation
As mentioned in Sect. 3.5.6, the beam modulation system was (schematic shown
in Fig. 4.1) used to purposely modulate the electron beam in a controlled way. The
goal was to provide a system that could produce roughly independent offsets in
76
each of the beam parameters, allowing for accurate determination of the detector
sensitivities on the scale of hours, as well as suppressing contributions from intrinsic
correlations in the beam. The system consisted of four pairs of air-core inductive
copper coils - two pair for X-like motion and two for Y-like motion - placed along the
beamline (Fig. 4.2). Each coil was driven by the output of a 16 bit VMIVME-4145
Waveform Generator. These waveform generators provided 4 channels of output per
board and have options for both internal or external triggering. For each channel the
sinusoidal waveform was built using a 64k word sample buffer. Once a waveform
of the desired frequency was lo added into the sample buffer, the external signal
was used to trigger the output on each desired channel. The signal output of each
channel was split, with one copy sent through a JLab-designed power amplifier card
(TRIM-I1) and used as a driving signal for the modulation coils. The other copy
was sent to an ADC, where it was used as both a diagnostic and as a reference signal
to lock the measured beam monitor response to the driving signal. The later point
became a powerful tool in the analysis methodology when dealing with the effects
of the fast-feedback system (FFB) on the beam monitor response to modulation.
Energy modulation was carried out in a similar fashion to position modulation,
with a driving signal being sent to the input of a superconducting radio-frequency
(SRF) cavity in the south linac of the accelerator. By modulating the input voltage
to the cavity we changed the electric field in the cavity and thereby the energy of
the electron beam. A final output channel was dedicated to producing a linearly-
increasing periodic function we defined as a “ramp” signal that was run directly
into an ADC. Since all outputs of the waveform generator were triggered via a
single external signal, the output wave-forms were synchronous. For every trigger
1Originally we had planned on using the second generation version of this power amplifier,however the engineering group at JLab found problems with them and they were removed.
77
that produced a driving signal the ramp channel was also triggered; through this we
were able to determine the phase of the driving signals.
FIG. 4.1: The modulation coil pairs are shown in red on the far right. Each coil wasdriven by an amplified signal from the VMIVME-4145 wave-form generators seen inthe center. A second copy of the driving signal was recorded by the main DAQ[11].
(a) Air-core coil used for drivenmodulation.
(b) Air-core coils in position on thebeamline.
FIG. 4.2: Copper air-core coils used in the modulation system.
During the design of the beam modulation instrumentation it was important to
determine whether the modulation hardware was capable of providing the required
field integrals, at the frequencies desired. Both the frequency and type of waveform
modulation were important to these tests. We wanted to maximize the up-time
of our measurement at maximum amplitude. Higher frequency modulation gives
78
a higher number of measurements per minute and using a waveform that spends
more time at maximum displacement makes each measurement more precise, e.g. a
square wave would be preferable over a sine wave. Bench tests were performed using
the air-core coils driven using a FANUC VME function generator with the TRIM-II
power amplifiers. During the bench test the coils were modulated at frequencies of
10 Hz – 500 Hz; a Hall probe was used to measure the field integral and the power
amplifier input voltage versus the coil current was recorded for different frequencies.
As a result of the bench top tests it was determined that the maximum reliable
operating conditions for the system was 250 Hz at a current of 3 A [62]. The choice
was made during Run I to operate the modulation system at a frequency of 125 Hz
- this is slightly off the poles of the FFB system which cancels beam noise at 60 Hz
and higher harmonics. This choice of sampling frequency provided an acceptable
waveform while providing sufficient livetime for each modulation cycle.
FIG. 4.3: An example beamline optics simulation to generate a 50 µm offset at thetarget. The red arrow shows the direction of the beam. The coils are representedby C1 and C2 and show the optimal locations along the beamline to apply kicks[12].
The position, angle, and energy of the beam at the target is determined by the
transfer function of the beam from the modulation coils to the target. By applying
79
the proper set of kicks to the beam at strategic locations along the beamline we were
able to generate the desired offsets at the target. The ratio and amplitude of the
kicks for each modulation type were determined using beam transport simulations
in OpTIM[63]. Taking advantage of the time reversal invariance of the electromag-
netic interaction, the beam was run backwards through the simulation, starting in
the desired position and moving upstream along the beamline. An example optics
simulation for an initial 50 µm offset at the target can be seen in Fig 4.3. Here the
beam moves from left to right, as indicated by the red arrow, propagating upstream
along the beamline. The simulated path shows the orbit, given the transport optics,
the beam would need to result in a pure position offset of 50 µm at the target.
The zero-crossings of the time reversed beam are important in the sense that they
provide a direct mapping of the position(angle) at the target to the beam angle at
the coil position. Placing kicker coils at these locations allowed us to generate the
forward orbit resulting in the pure position(angle) offset we desired. Similar simu-
lations were carried out for each type of modulation; these simulations determined
the starting currents and coil ratios that would be used during running. One of
the default running configuration can be found in Table 4.1. In the table you can
see the coil currents required to produce the given offsets; one important thing to
notice is the near degeneracy in the Y-like coil currents. The basic concern here
was that if two or more beam parameters - which make up our basis set - were not
sufficiently independent there could be difficulty in extracting the sensitivities. This
was an ever present concern throughout the extent of the experimental running, and
one that was never completely resolved. During times that a change in accelerator
optics was suspected, plots of the beam trajectory were made by plotting the BPM
response to modulation to each kind of modulation as a function of Z position along
the beamline. These optics plots provided a snapshot of the current beam optics,
TABLE 4.1: Example of a standard run configuration from Run I modulation. Thesenumbers were adjusted throughout the experiment as the beam optics changed.
and provided a convenient way to investigate any position-angle mixing that might
have been present. In order to address these optics changes and try to optimize the
response, special tune-scan runs were taken in which the coil tune would be varied
by ±10%. The resulting change in the optics was then studied and a “best” tune
value was extracted.
The controls system for the beam modulation system was written in State Nota-
tion Language (SNL) using sequencers. SNL is a powerful programming framework
that is useful in real-time control systems; the run-time sequencer drives the controls
system into different states based on different events and relieves the complexity of
task scheduling and event handling that is a staple of a real-time multi-tasking
environment. Both SNL and the sequencer are components of a larger controls
framework, the Experimental Physics and Industrial Controls System (EPICS2).
EPICS is an interactive development toolkit and real-time controls environment for
physics and industrial applications. The sequencer setup was especially important
to the modulation system as it allowed monitoring of the modulation related vari-
ables in a simple way as well as scheduling of different modulation cycles. Once
initialized, the controls system would read in the desired modulation setup from a
user defined configuration file - this configuration file would include which channels
were to be initialized, the frequency at which to modulate, and the driving signal
2http://www.aps.anl.gov/epics/
81
amplitude defined in amps. Once the configuration files had been read-in, the wave-
form generator boards were transitioned into a ready state. An external trigger was
then applied to the boards; this signal would cause the channels that were initialized
for a certain modulation type to trigger and run through a full set of cycles. Once
the modulation cycles were finished, the system would set all channels into an OFF
state and move to a wait state until it was time to wake up and modulate again.
The modulation system ran continuously during most of the production running,
stepping through a modulation sequence in each beam parameter; a set of micro-
cycles containing one instance of modulation in each beam parameter makes up a
macro-cycle. During each modulation cycle a pattern number flag was set in the
DAQ; there was a unique pattern number for modulation in each beam parameter,
which allowed us to identify the type of modulation of each micro-cycle during anal-
ysis. The modulation sequence consisted of driving each pair of modulation coils
in micro-cycles of about 4 seconds, or 512 cycles at 125 Hz. There was a down
time of ≈ 75 seconds between each micro-cycle of modulation which was needed to
reconfigure the boards for the next modulation type. A full modulation macro-cycle
was 320 seconds, after which time the modulation system would reconfigure and
repeat the macro cycle; the modulation data composed about 5% of our production
running.
4.1.3 Methodology
As mentioned previously, measurement and correction of false asymmetry caused
by helicity-correlated beam systematics was done using two methods: using standard
linear regression technique applied to natural beam jitter, and using large-amplitude
driven beam motion via the beam modulation system. The correction to the raw
82
FIG. 4.4: Schematic of beam modulation cycle timing. Each pulse section is 512cycles of sinusoidal function at frequency of 125 Hz. Pulse are broken down intomicro-cycles that make up a full macro-cycle.
asymmetry due to helicity-correlated beam systematics was determined using the
beam modulation system outlined in the previous section. The raw asymmetry was
corrected for false asymmetries according to
Acorrected = Araw − Afalse (4.1)
where Afalse is the correction due to false asymmetries and is given by
Afalse =1
〈Y 〉∑i=1...5
∂Y
∂Xi
∆X. (4.2)
Here the detector sensitivities are given by ∂Y/∂Xi and represent the direct correlation
of the detector yield to changes in a given beam parameter. It is important to note
that this is not directly measurable; the beam modulation strove to provide linearly
independent modulations of the beam but could not completely remove cross cor-
relations between beam parameters, e.g. position modulations also had components
of energy and angle mixed in. Extraction of the detector sensitivities was done by
83
considering the response of the detector yield with respect to the modulation signal;
the modulation signal is analogous the modulation type. Expanding with respect
to the beam parameters,
∂Y
∂Ci=∑
j=1...N
∂Y
∂Xj
∂Xj
∂Ci, (4.3)
where Cj represents a time-dependent signal synced with the modulation driving
signal, yields the detector sensitivities. The beam monitor matrix, given by ∂Xj/∂Ci,
gives the correlation of the beam parameters measured in the monitors to each
modulation type. Both the beam parameter correlation matrix and the detector
response to each modulation type are directly measurable. The detector sensitivities
can then be extracted by matrix inversion:
∂Y
∂Xj
=∂Y
∂Ci
[∂Xj
∂Ci
]−1
. (4.4)
This can be written in matrix form as:
MS = MCM−1M (4.5)
As was explained in Sec. 4.1.2, the beam modulation system used pairs of coils,
driven with different amplitudes, which produced a composite response in the beam.
While the individual driving signals were available in the DAQ, there was no read-
back of the composite driving signal. Instead the ramp signal, which was synced to
the coil driving signals and common to each modulation type, was used. Refering
to Eq. 4.3 above, Cj is a parametrized version of the ramp signal given by
C(t)i = sin(αr(t)− φ), (4.6)
84
where r(t) is the ramp function. Parametrizing the response in this way allowed
us to pick out only the component of the response directly proportional to the
driving signal, which was important in circumventing components in the response
such as Fast-Feedback (FFB) system, discussed later in this chapter. Using the
detector response in Eq. 4.4, the sensitivities to each modulation type (MC) were
determined using a fit to a linear regression line,
M jC =
∑i=1...N
(Cji − 〈Cj〉)(Y j
i − 〈Y j〉)∑i=1...N
(Cji − 〈Cj〉)2
, (4.7)
where the sum is over the N modulation events in a micro-cycle, and j is the mod-
ulation type. The correlation matrix (MM) was calculated in a similar fashion for
each modulation type and beam monitor as,
M jkM =
∑i=1...N
(Cjki − 〈Cjk〉)(M jk
i − 〈M jk〉)∑i=1...N
(Cjki − 〈Cjk〉)2
. (4.8)
Here j represents the modulation type, and k is the beam parameter. This was
done for each micro-cycle in a given run and the error weighted average of MC and
MM were used in Eq. 4.5 to complete MC and MM respectively. Given that MM
is invertible, i.e. |MM| 6= 0, the detector sensitivities to each beam parameter (MS)
can be extracted as shown in Eq. 4.5 – 4.6.
One of the complications that came about when syncing to the ramp signal was
the way in which the ramp signal was sampled. The DAQ samples signals coming
into the ADCs at 960 Hz, and given the modulation frequency of 125 Hz, there are
∼8 data points for each ramp cycle. The issue comes about at the transition point
between ramp cycles; when the ramp signal transition falls in the center of an ADC
sample window, it creates a point that is not consistent with the functional form
85
given by r(t). In Fig. 4.5 a sample output of the ramp signal is shown; depending
on how close to the center of the ADC window the transition point falls the ramp
signal coincides with a different phase. The results of this can be seen by plotting
one of the modulation driving signals versus ramp, as can be seen in Fig. 4.6. The
seemingly linear function that tracks through the center of the driving signal is
due to the ramp transition points falling in progressing different places in the ADC
window. In order to remove this effect, a simple linearity test was applied in the
modulation analyzer. This test takes advantage of the fact that for each ramp point
in the ADC the four subblocks used to build that point are also saved. Using the
was used. The size of δr was determined analytically to be 50, which removed
the majority of the transition points while not being so strict a cut as to remove
large amounts of data. Referring again to Fig. 4.6, the results of the above cut
can be seen as the points in black. The points removed by this cut become an
important systematic in the determination of the detector sensitivities; by removing
these points, a phase gap is created in the modulation data. The details of how this
effects the determination of the detector sensitivities can be found in Sec. 4.1.4.
In later analyses a better way to deal with this effect was developed which actually
reassigned the points calculated at the transition point to where they are expected in
the phase gap according to the linear nature of the ramp function[64]. This method
was not applied in this analysis and won’t be discussed further.
Early in the experiment, during commissioning of the modulation system, it
was decided that unlike previous implementations of similar systems[65, 66], FFB
86
FIG. 4.5: The ramp signal read out by the ADCs. Because the DAQ samples at arate of 980 Hz and the ramp signal has a frequency of 125 Hz, the ADC records ∼8 points per cycle. When the ramp transition falls in the middle of an ADC samplewindow the ramp point recorded does not match up with what is expected at thattime; see data surrounded by green vertical lines.
FIG. 4.6: An example plot of the modulation driving signal for one coil plottedversus the ramp function is shown in red. The edge effects of the ramp signal beingsampled by the ADC point recorded at an incorrect phase form a linear functionthat tracks through the center of the driving signal. Using a simple linearity cutmany of these points (shown in black) can be removed.
87
FIG. 4.7: The phase of the BPM response is shown as a function of position along thebeamline. The dashed blue lines indicate the position of the vertical and horizontalmodulation coils.
would be enabled during running, except during energy modulation. This was due to
fears that disabling the FFB system, which actively monitored and tried to correct
beam noise, would introduce significant noise into the experimental measurement.
This decision came following a test run of the modulation system which showed two
things [67]: the energy FFB completely flattened the energy modulation signal and
the suppression of the position modulation was small (O(5%)). Unfortunately, the
effect on the phase of the BPM response to modulation was not recognized during
the commissioning tests.
In Fig. 4.7 the phase of the BPM response in each beam monitor along the
beamline is shown; of primary importance are the final five BPMs before the target
which are used to construct the “target” variables and therefore directly affect the
extraction of the detector sensitivities. To understand the phase shift due to FFB
we looked at how the FFB algorithm monitored and made corrections to the beam
position. Along the beamline certain BPMs were monitored by the FFB system.
88
These feedback BPMs were used by the FFB system, which samples the position
noise over two second windows at 1.8 kHz, and creates a waveform to null out
horizontal (vertical) displacement with an equal but opposite position offset. This
is equivalent to another modulation coil that has a delayed response - or a different
phase - to the modulation. The total response of the BPM to modulation can be
parametrized as the sum of two harmonic functions[68] with different phases
X(t) = A sin(ωt+ φmod) +B sin(ωt+ φffb). (4.10)
This can be be rewritten as
X(t) = A′ sin(ωt+ α), (4.11)
where
α = tan−1 A sinφmod +B sinφffb
A cosφmod +B cosφffb
(4.12)
A′ =√
(A sinφmod +B sinφffb)2 + (A cosφmod +B cosφffb)2. (4.13)
Here φffb is the response of the FFB to the modulation signal and φmod is the
composite response of the beam to modulation. The problem this presents is that the
phase of the composite response is a function of the relative phases and amplitudes of
the modulation coils and the FFB system. The amplitude response of a given BPM
to modulation depends on the transfer function from the coils to a particular point
along the beamline. This means we had a position-along-the-beamline dependent
phase response to modulation; another way of saying this is that we are spanning
more than one dimension in phase space. The solution to this was that we measured
the response to modulation with a fixed phase. Considering the response to a given
89
modulation as the sum of harmonic functions, we have
X(t) =∑i=0...N
xi sin(ωt+ φ) = xffb+drives sin(ωt+ φ) + xffb
c sin(ωt+ φ) (4.14)
where xffb+drive is the component of the response due to the driving signal plus the
sine component of the FFB, and xffbc is the cosine component of the FFB response. If
we consider only the sine component of the BPM response, the FFB-only component
averages out, leaving us with an amplitude that is proportional to the driving signal.
In fact, because the two harmonic functions are time-orthogonal, the FFB compo-
nent will always cancel out, allowing us to use any arbitrary phase. For the purpose
of the analysis presented here, φ was chosen to be zero. This corresponds to φ = 0 in
Eq. 4.6. This postulate rests completely on the assumption that the FFB response
is not dynamic on the time-scale of the modulation. A detailed presentation of how
this phase lock postulate was tested is presented in Sec. 4.1.4.
4.1.4 Modulation Regression Analysis
The previous section explained the basic methodology used, as well as some of
the difficulties encountered, in the measurement of the detector sensitivities using the
beam modulation system. This section details the diagnostics used to ascertain the
quality of the extracted sensitivities and the subsequent corrections to the measured
asymmetry. In Sec. 4.1.3 it was explained that due to the FFB being left on
during modulation, there was a phase shift induced into the BPM response used to
calculate the detector sensitivities. The proposed solution was to model the total
BPM response as the superposition of a set of harmonic functions, which could
be separated into components proportional to the driving signal plus part of the
FFB, and a time-orthogonal piece proportional only to the FFB signal. This led
90
to the postulate that we could choose any arbitrary phase to extract the detector
sensitivities. To test this, four different “sets” were chosen - each with a different set
of phases - to extract the detector sensitivities. Shown in Table 4.2 are the different
TABLE 4.3: Wien average sensitivities for each modulation set. Positions sensitiv-ities are in units of pm/mm and angle sensitivities are in units of ppm/µrad.
91
was significant dispersion in the beam-tune, increasing the correlation between the
position and energy of the beam; this correlation changed as the beam tune drifted.
In this model, the definition of X, X′, and E changes and therefore a slight shift in
the extracted sensitivities is expected. In Fig. 4.8 - 4.12 you can see correlation plots
of the virtual target BPMs and bpm3c12X which demonstrate the strength sharing
present in the extracted sensitivities. Each plot shows the set of Run I modulation
extracted sensitivities on the runlet basis3 for each monitor.
One likely cause of the correlation of the extracted sensitivities was correlated
noise in the monitors used to form the virtual BPMs as well as bpm3c12X; with this
in mind we expect the correlation to be suppressed when averaging over longer time
scales. Correlation plots of the Slug-averaged sensitivities are shown in Fig. 4.13 -
4.17, and show that the correlation between the extracted sensitivities is suppressed
as the monitor noise is averaged out, as expected.
3In this case runlet refers to segments of a full run, each segment making up approximately 6minutes of data.
92
FIG. 4.8: Correlation between TargetX and TargetX′ sensitivities.
FIG. 4.9: Correlation between BPM3c12X and TargetX sensitivities.
93
FIG. 4.10: Correlation between BPM3c12X and TargetX′ sensitivities.
FIG. 4.11: Correlation between TargetY and TargetX sensitivities.
94
FIG. 4.12: Correlation between TargetY and TargetY′ sensitivities.
FIG. 4.13: Slug average correlation between TargetX and TargetX′ sensitivities.
95
FIG. 4.14: Slug average correlation between BPM3c12X and TargetX sensitivities.
FIG. 4.15: Slug average correlation between BPM3c12X and TargetX′ sensitivities.
96
FIG. 4.16: Slug average correlation between TargetY and TargetX sensitivities.
FIG. 4.17: Slug average correlation between TargetY and TargetX′ sensitivities.
During both Run I and Run II run periods data was taken with the FFB system
97
deactivated. The original purpose of this, during Run I, was to study the effect of
the FFB system on the response amplitude during modulation; if FFB significantly
suppressed the BPM response amplitude we could not run it during modulation. The
result of the Run I test was that the FFB reduced the BPM response ∼5% and there
was a slight amplification of the Y response. Unfortunately, as mentioned previously,
the phase shift due to FFB interacting with the modulation was overlooked. The
response to modulation for each BPM during both runs with and without FFB
active can be found in Fig. 4.18-4.22. From the plots the effect of the FFB on the
BPM response is clear, with the effect on the Y-like BPMs being most significant.
It is important to note that the phase shift of the BPM response during the FFB
off run should be insignificant; in the case of energy modulation, because the energy
lock is always disabled, there should be no phase shift in either case. The results
for the FFB off run match expectations with the exception of TargetY′ modulation.
Referring to Fig. 4.22, the phase shift seen in TargetY′ with FFB off is ∼ 70- 80.
The fact that there is a significant phase shift in the BPM response to modulation
when the FFB system is disabled, is suggestive of there being another set of coils
coupling to the modulation. This was looked into more thoroughly by the beam
corrections group at the University of Virginia; currently no source for this coupling
has been found. One thing that group noticed was noticed was that the phase shift
with FFB turned off is not present in the Run II results. The analysis to determine
what the issue was is currently on-going, however, as mentioned above, the addition
of an additional phase should not matter in the analysis as long as the shift does
not change with time, however, it is still desirable to understand the source.
98
FIG. 4.18: TargetX response to modulation during runs with and without FFBactive.
FIG. 4.19: TargetX′ response to modulation during runs with and without FFBactive.
99
FIG. 4.20: BPM3c12X response to modulation during runs with and without FFBactive.
FIG. 4.21: TargetY response to modulation during runs with and without FFBactive.
100
FIG. 4.22: TargetY′ response to modulation during runs with and without FFBactive.
FIG. 4.24: The same as the above plot but profiled. The pull due to the cut rampdata is much more pronounced when looking at the average versus R(r(t)).
One of the issues that arose from the BPM response phase shift was, when
101
FIG. 4.23: The response of the TargetX BPM to X modulation plotted versusR(r(t)) = sin( π
180r(t) + φ), where r(t) is the ramp function. Because of the phase
shift from the FFB system the response traces an ellipse in phase space; near thecenter of the ellipse the missing data due to the ramp cut can be seen. This missingdata and the ellipsoid nature of the response-induced extra error into the sensitivitiesextraction and made analytical calculation of the errors on the sensitivities moredifficult.
looking at the correlation between the BPM response and the sine of the ramp
function (recall Eq. 4.8), that the result was not linear. Looking at the plot of
the TargetX virtual BPM versus the sine of the ramp (Fig. 4.23) - defined as
R(r(t)) = sin( π180r(t) + φ), where r(t) is the ramp function - you can see that the
result traces out an ellipse in phase space. This causes problems in two ways. First,
because the ramp cut removes events calculated incorrectly during the transition
period of the ramp function, the ellipsoid is not filled continuously in phase space,
i.e. the ellipse is double valued in places and single valued in others. Fig. 4.23 shows
a scatter plot of TargetX versus R(r(t)) and Fig. 4.24 shows a profiled version of the
same plot. The data removed by the ramp cut causes the response ellipse (4.23) to
102
be non-continuous, potentially causing the calculation of the correlation performed
in Eq. 4.8 to be incorrect. The second issue, which would be present even if the
ramp cut was not applied, is that the error on the correlation calculation will be
incorrect; this is because we are applying a linear model to a non-linear response.
Following the completion of the modulation analysis replay, the effect of the
missing ramp data was investigated. A method to fill in the missing ramp data
and make the response continuous was developed at the University of Virginia; this
method of filling in the missing ramp data would have required a new full replay
of the data and was not used in the present analysis, however, early results using
the ramp fill method were compared [69] to the results found here and showed only
a small change. Given the difficulties in computing the errors on each extracted
sensitivity explained above, the errors were instead extracted by looking at the
RMS/√N for each sensitivity on the Slug and Wien level. Shifting the analysis to
the Slug and Wien levels not only allowed for the extraction of more accurate error
bars, but Slug or Wien level sensitivities facilitated the use of more accurate detector
sensitivities; the sensitivity of the detectors to changes in the beam parameters was
a property of the apparatus and should have been constant. These Slug or Wien
level sensitivities can then be applied to the data for more accurate corrections.
Looking at the single detector sensitivities as a function of octant number pro-
vides an important diagnostic tool as well as providing insight into the nature of
the detector sensitivities. In Fig. 4.25 - 4.27 the Run I average single detector
sensitivities are shown as a function of octant number. Recall from Sec. 3.5.4, the
main detector is an azimuthally symmetric array of quartz Cerenkov detectors. The
octants are numbered as shown in Fig. 3.12, with octants 1 and 5 in the horizontal
plane. With these definitions, it is expected that the measured detector sensitivity
for each bar will be characteristic of the modulation type.
103
For instance, for an X-like modulation the single detector sensitivity as a func-
tion of octant number should be sinusoidally varying with peaks at octant 1 and
octant 5; these octants are in the plane of modulation and should be most sensitive.
In Fig. 4.25 a plot of the Wien-average single detector X-sensitivity as a function of
octant number is shown. The detectors in the plane of modulation show the great-
est sensitivity to X-like modulation. Taking note of the size of the largest single
detector sensitivities, which are on the O(±5500 ppm/mm), we get a factor of ∼3
suppression in the octant-average value due to the detector symmetry; the suppres-
sion in Y is much greater, ∼24. The size of the detector averaged sensitivities could
be attributed to some degree to the beam not being centred on the neutral axis of
the detector; if the parts of the scattered electron profile fall off the detector bar it
would increase the position sensitivity. Studies of the relative main detector widths
[70] do suggest that our choice of average beam position was not optimal. Another
likely source was background caused by beam halo. Beam halo is a low-density
collection of particles that gather around the core of the electron beam. Halo can
be caused by a number of things including: single and multi-particle scattering, ma-
chine non-linearities, and noise in the injector. During beam modulation the beam
halo can interact with elements along the beam line causing backgrounds which
could manifest as a monopole in the detectors.
104
FIG. 4.25: The Run I single detector X position sensitivities over Run I versus oc-tant number for X modulation. Because X modulation is in the horizontal plane,detectors 1 and 5 are most sensitive to this modulation type. The sinusoidal vari-ance of the detector sensitivities, with peaks at 1 and 5, is characteristic of X-likemodulation. Error bars are included but are smaller than the markers.
FIG. 4.26: The Run I single detector Y position sensitivities over Run I versusoctant number for X modulation. Because Y modulation is in the vertical plane,detectors 3 and 7 are most sensitive to this modulation type. Error bars are includedbut are smaller than the markers.
105
FIG. 4.27: The Run I single detector energy sensitivities over Run I versus octantnumber for E modulation. Ideally, a change in energy should be seen as a change inQ2 and effect each detector uniformly. The expected response would be a detectormonopole. Error bars are included but are smaller than the markers.
The extracted detector sensitivities were studied at the runlet scale - “runlet”
for modulation data is a different unit than mentioned above and corresponds to
approximately half a full run - slug, and Wien levels. The detector sensitivities, if
properly determined, should be a property of the detector apparatus and therefore
should be static. Studying the detector sensitivities at longer time scales allows
us to reduce the statistical noise and determine both the stability of the extracted
sensitivities and the time scale on which it is most appropriate to apply corrections to
the asymmetry. Shown in Fig. 4.28 - 4.32 are the Slug-average detector sensitivities
for Run I, weighted by the main detector error. The extracted detector sensitivities
for each beam parameter on the Wien level are shown in Table 4.7. Aside from the
occasional outlier, the Slug-level sensitivities are generally stable. One aspect where
there does seem to be a shift in the sensitivities is bpm3c12X; starting after Slug
106
100 the sensitivity starts to drift.
FIG. 4.28: Run I Slug-average Set 1 X sensitivities.
FIG. 4.29: Run I Slug-average Set 1 X′ sensitivities.
107
FIG. 4.30: Run I Slug-average Set 1 BPM3c12X sensitivities.
FIG. 4.31: Run I Slug-average Set 1 Y sensitivities.
108
FIG. 4.32: Run I Slug-average Set 1 Y′ sensitivities.
A suitable metric for whether the modulation set results agree between phase
sets is the Wien-level total corrections to the asymmetry for each set. In Table 4.4
the total correction for each Wien and each set is shown. It should be noted that
because the position differences are different from Wien-to-Wien there is no reason
for the corrections to be the same for different Wiens. From the table it can be seen
that, within errors, the corrections are consistent between sets. This shows that the
choice of phase, while changing the strength sharing, does not matter at the current
level of precision in extraction of the sensitivities. For the analysis presented in this
thesis, Set 1 was chosen. This choice was dictated by the simplicity of having no
phase shift and by the size of the extracted errors on the sensitivities.
One of the issues with determining the quality of the beam modulation sensi-
tivities comes from the fact that the data used to extract the detector sensitivities
is not the same data that is corrected; this is not a problem for the standard linear
regression of natural beam motion as it uses identical data sets. If we are correctly
extracting the detector sensitivities and they are not changing there should be no
Run I Avg. -1673.8 ± 7.7 81.8 ± 0.4 -1480.6 ± 1.4 264.1 ± 5.5 -1.1 ± 0.3
TABLE 4.7: Wien-level sensitivities for the main detector system. The Run I valuerepresents the error weighted average of each Wien-level result. Positions are givenin ppm/mm and angles are given in ppm/µrad.
it is mathematically required that the widths are reduced. This is not the case for
the modulation-corrected data. While the modulation corrections are expected to
reduce the raw width, it is not a mathematical requirement since the modulation
extracted sensitivities are determined from an independent set of data. Below in
Fig. 4.43 the main detector average corrected widths, for both beam modulation
and linear regression Set 11 corrected asymmetries, are shown. The Set 11 linear
regression set uses the same independent variables as the beam modulation. The
beam modulation corrections were done using Wien-averaged sensitivities and the
LRB corrections were done using both quartet and the Wien-average sensitivities.
Both the linear regression and the modulation corrections are shown to reduce
the asymmetry width. The greatest width reduction comes from the quartet-level
linear regression corrections, but all methods had a positive effect on the asymmetry
118
FIG. 4.43: The slug average detector widths are shown for raw, corrected via mod-ulation, LRB, and wien average LRB for slugs 42- 59. The maximum reduction inwidth is achieved using the quartet level LRB sensitivities.
width. The fact that the width reduction due to the Wien-average LRB sensitivities
is worse than the quartet level corrected widths is suggestive of a hidden variable.
An argument could be made that the sensitivities are changing with time and there-
fore the Wien average corrections are wrong, however this seems unlikely given the
stability of the beam-modulation extracted sensitivities. As a note, changing values
of the slug average widths are expected due to changing beam properties during
each slug; a run period such as slug 49 could simply be due to a section of time with
unusually large position differences. The average beam current during Run 1 was
156 µA and the main detector current weighted rate was ∼4.4 MHz/µA[58]. This
implies that the combined detector width due to counting statistics alone should be
O(209) ppm per 4 ms quartet. The remaining noise is likely due to other sources
of noise such as beamline backgrounds, detector energy resolution, BCM resolution,
TABLE 4.9: The residual correlations of the main detector to each degree of freedomin the beam after beam modulation corrections. Each corrections was determinedand is listed as the Wien average value. Position correlations are given in ppb/nmand angular correlations are given in ppb/nrad.
4.1.6 Average Position Differences
The Run I Wien-Average helicity-correlated position differences for each of the
beam parameters are given in Table 4.8. Each beam parameter is weighted by the
main detector error using the prescription found in Appendix A.1. Calculation of
the Wien-average position differences was subject to a data quality cut given by
|xi − 〈x〉| < nσ√σ2xi
+ σ2〈x〉 (4.17)
where xi is the position difference for each runlet, 〈x〉 is the detector-weighted Wien-
average position difference before cuts, σ2xi
is the runlet error, and σ2〈x〉 is the error on
the Wien average position difference. The size of the cut was defined by nσ, which
was the number of sigma from the mean that was acceptable. For the purpose of
this analysis, nσ was set to be six; a six sigma cut removed the major outliers in the
data while leaving most of the data intact. The position differences are also subject
to a number of beam stability cuts inside the standard analyser as well as cuts on
whether there were failures in any of the linear regression schemes. Shown in Table
4.10 are the number of events lost for different cuts. The number of events lost due
to more stringent cuts is relatively small up to 6σ.
The Wien 0 result is used in this analysis due to the unavailability of the Run 1
results.
FIG. 5.1: Regressed aluminium asymmetry plotted versus slug for both IN and OUThalf-wave plate settings. This does not include corrections due to polarization orbackgrounds.
5.2.3 Aluminium Dilution
The dilution of the measured yield due to electron scattering from the alu-
minium windows was measured using two techniques: evacuated cell method and
gas extraction method. The easiest of the two methods is the former in which the
target cell is evacuated and the scattering signal is recorded. The cell is then refilled
and another measurement is taken. In the full condition, the liquid hydrogen works
to cool the target windows. Because the liquid hydrogen is needed for cooling, one
concern with the evacuated method was potential thermal and structural damage
131
to the windows. Because of this the measurements were done at much lower cur-
rents: 0.2 - 1.0 µA. One thing to note is that there were several hours between the
evacuated and full measurements due, in part, to the time required to refill and cool
the target. In order to eliminate potential errors in the measurement due to time
dependent changes, and BCM calibration uncertainties, a reference target, the yield
from which should not change over time, was used to normalize both measurements:
Rnormempty =
Rempty
Rreference
, Rnormfull =
Rfull
Rreference
(5.13)
A 0.5% radiation length carbon target, located on the DS target ladder, was used
as a reference. The dilution factor is the ratio of Rnormempty/Rnorm
full . As before, con-
tributions from the US and DS windows must be treated separately. The process
of applying the radiative corrections is guided by studies of aluminium targets of
different thickness and simulations. The Run I dilution factor can be seen in Fig.5.2
for opposing octants. A discrepancy of ∼2.4% can be seen between the normalized
and unnormalized measurements which is attributed to the difference in reference
target yields [77]. The final extracted dilution, including corrections [78], is given
The gas extraction method uses gaseous hydrogen in the target at different pres-
sures to determine the signal in the detectors versus pressure. By extrapolating to
zero pressure, the empty target signal can be determined, allowing for the extraction
of the dilution factor. While this method of measurement serves as a check on the
evacuated method, it is not the cleanest method of extraction. Uncertainties such
132
FIG. 5.2: The Run I Dilution for opposing octants is shown for both normalizedand unnormalized cases. The discrepancy between with and without normalisationis attributed to a difference in the reference target yields.
as local density changes due to heating, electronic deadtime, radiative corrections
due to the addition of hydrogen to the target cell, and BCM linearity between the
gas and full states make this not the optimal approach.
5.2.4 Beamline Background
The beamline background encompasses low energy events hitting the main de-
tectors not originating from elastic scattering in the target. Possible sources in-
clude: showers from the collimator and shield wall, octant-to-octant cross-talk, and
beamline scattering. These background sources, while somewhat smaller than the
Aluminium target window background, are an important contribution to the mea-
sured asymmetry. In general, much of this is suppressed via shielding and the lead
pre-radiators installed directly in front of the main detectors and the tungsten plug
133
which was installed in the downstream side of the first collimator. The pre-radiators
are effective at suppressing low energy events entering the detectors while also in-
creasing the size of the elastic events from the target.
5.2.5 Beamline Background Dilution
Measurement of the beamline background dilution was done using a blocked
octant technique. In order to eliminate line-of-site from the target events, 3” of lead
was installed directly downstream of the defining collimator in octant 7. Blocking
the octant allowed for a measurement of background dilution as the ratio of the
blocked signal to the unblocked total signal. Blocked octant measurements were
taken in May 2011 at two QTOR currents, 8921 A and 6700 A (inelastic), and
on both the lH2 and 4% DS aluminium targets. A more detailed discussion of
the blocked octant study can be found in K. Meyers thesis [8] and Wade Duvall’s
analysis summary[79]. The results of the blocked octant study give an upper bound
on the beamline dilution of
fb3 = 0.00218± 0.00064. (5.15)
The average of the blocked octant data, including good and bad halo periods, was
used in previous analyses. Due to the increased correlation between the main detec-
tors and the background monitors described below it was decided that a conservative
estimate on the Run I dilution was more appropriate. The value used is based on
values extracted from bad halo data. This is expected to be an upper bound due
to the fact that significant increases would mean significantly more beam being
dumped into the tungsten plug, accompanied by increased temperatures in the plug
readings, and frequent beam trips[80]. This is not something that was observed.
134
Beamline Background Asymmetry
The asymmetry contribution due to beamline background sources was deter-
mined using results from the blocked octant study [79], however it was not done
directly. The blocked octant study was completed near the end of Run II data tak-
ing and there was no analogue in Run I. Because of this and the fact that the MD9
background detector was moved during Run II the USLUMIs are used as a common
element between the runs to indirectly measure the beamline background outside
of the blocked octant study. The details of how this is done can be found in Mark
Pitt’s log entry [81]. According to the blocked octant study analysis the following
relationships can be derived.
AMDunblocked = 0.0018× AMD
blocked (5.16a)
AMD9unblocked = 0.094× AMD9
blocked (5.16b)
AMDunblocked = (
fMD
fMD9
)(AMD
blocked
AMD9blocked
)AMD9unblocked (5.16c)
AMDunblocked = 0.044× AMD9
unblocked (5.16d)
Here the coefficient in Eq.5.16b and 5.16c gives the measured dilution factor and
Eq.5.16d gives the relation between the unblocked detector signals in terms of the
dilution factors and the measured blocked octant values. Physically, Eq.5.16d gives
the ratio of the background asymmetry contributions in terms of MD and MD9.
This is tied to the USLUMI by looking at the ratio of the asymmetries between
MD9 and the USLUMI during Run I.
AMD9unblocked = 0.626× AUSLUMI
unblocked (5.17)
135
Putting Eq. 5.16d and 5.17 together we get
AMDunblocked = (0.0276± 0.0276)× AUSLUMI
unblocked. (5.18)
An error bar of 100% has been assigned due to uncertainties in the fraction of the
signal seen in MD9 that is from beamline background. Using this relationship the
average beamline asymmetry was determined to be
Ab3 = 10.94± 10.95 ppb. (5.19)
5.2.6 Inelastic Background
The inelastic background is from inelastic scattering events, dominantly from
the nuclear N → ∆ transition, which fall into the experimental acceptance of the
main detectors. The inelastic asymmetry was expected to be O(10) larger that
than the elastic asymmetry, however the fraction of inelastic events in the detectors
to that of the elastic events makes the total contribution to the total asymmetry
relatively small. Measurement of the inelastic asymmetry was done by changing
the QTOR current to a value that maximized the inelastic signal in the detectors;
according to simulation (Fig. 5.3) this was 6700 A. By changing the QTOR current
from 8921 A to 6700 A the fractional contribution of inelastic to elastic events was
increased O(100) [13]. The inelastic asymmetry was found to be [13]
Ab4 = 3.02± 0.97 ppm. (5.20)
136
FIG. 5.3: Inelastic dilution from simulation as a function of QTOR setting[13].
The dilution due to inelastic events was determined using simulation[82] to be
fb4 = 0.0002± 0.0002. (5.21)
Results from this study found the simulated rates to be 10% below those determined
at the inelastic peak from data. Due to this, a preliminary 100% error bar was
assigned pending further analysis.
5.2.7 Total Neutral Background
The neutral background originates from secondary photons and scattering of
primary electrons from the collimators and shield walls; a small percentage of the
background is made up of pions and neutrons. Estimation of the neutral background
was done using the RIII chambers in counting mode; counting mode allows for
137
the detection of single events. Neutral events detected in the RIII chambers pass
through both the trigger scintillator and the main detectors. Both of these are
relatively insensitive to neutral events, however interactions such as neutron capture
and Compton scattering, while being below the threshold of detection in the trigger
scintillator, may shower in the pre-radiator or interact in the fused silica and cause
a signal in the MDs. These events can be classified by the amount of light they
produce. In this way, the neutral background fraction is defined by
nf =MDneutral × 〈Yneutral〉MDtotal × 〈Ytotal〉
. (5.22)
Here MDneutral is the number of detected neutral tracks and Yneutral is the light yield
from those tracks. The neutral fraction is formed as the ratio of this product for
neutral tracks versus all tracks. This analysis was done for each detector octant at
nominal QTOR current. An in-depth look at this measurement can be found in [5].
TABLE 6.1: The asymmetry and associated dilution for each background source.
after correction using beam modulation and polarization, the size of the background
corrections, and the final corrected asymmetry. The corrections were applied to the
measured asymmetry after modulation correction according to
Aphysics = Rtotal
Amsr
P−
n∑i=1
fiAi
(1−n∑i=1
fi)(6.1)
where fi and Ai are the dilution and asymmetries for each background respectively
and Rtotal is the total experimental bias correction. The asymmetry, Amsr, represents
the beam modulation corrected asymmetry and the beam polarization is given by
P.
The systematic error contribution from each background is calculated using
standard error propagation methods and is shown in detail in Appendix A.3. A
breakdown of the error contribution due to each correction, in terms of systematic
and statistical uncertainty, is shown in Table 6.2. The corrections are dominated by
Aluminium target windows, followed by polarization, and modulation corrections.
Corrections due to other background sources are relatively small.
The blinding factor addition to the error of the final asymmetry was computed
according to Appendix A.3 to be 68.52 ppb. The blinding factor, in this form,
represents a uniform distribution of error, however in order to fold it into the total
error on the asymmetry it needs to be rewritten in terms of a normal distribution.
146
FIG. 6.1: The contribution from various corrections to the measured asymmetry.The raw asymmetry is given on the left in orange. The corrections from beammodulation and background corrections are shown in blue, and the final correctedasymmetry is shown in purple on the far right.
Computing the second moment of the blinding factor gives a value of 39.56 ppb,
which can be added in quadrature with the systematic and statistical errors. Using
Amsr above as an input to Eq. 6.1 the final physics asymmetry including blinding
TABLE 6.2: Breakdown of the systematic errors going into the final uncertainty.
electromagnetic form factors are well determined in terms of global fits [93], which
is not the case for the electroweak form factors. Instead, as is shown in Sec.2.2.3,
the electroweak form factors can be written in terms of the quark electromagnetic
form factors,
Gp,ZE = 2(2Cu1 + Cd1)Gp,γ
E + 2(Cu1 + 2Cd1)Gn,γE + ε
(0)V Gs
E (6.3)
where ε(0)V encompasses contributions from charge symmetry violation [94]. Rewrit-
ing the asymmetry in this way, the global fit becomes a function of five independent
parameters: the weak vector couplings (C1u and C1d), the strange charge radius ρs,
the strange magnetic moment µs, and the axial form factors GZ(T=1)A and G
Z(T=0)A .
The strange quark form factors are given by [95].
GsE = ρsQ
2GD, GsM = µsGD (6.4)
148
FIG. 6.2: The contributions to the final error bar.
where
GD =1
(1 + Q2
Λ2 ). (6.5)
This parametrization was chosen to match the Q2 dependence experimental data
and the mass scale (Λ2) was set to 1 (GeV/c)2 to match PVES data up to Q2 =
0.63 (GeV/c)2 [36]. The axial form factors, separated into isovector (GZ(T=1)A ) and
isoscalar (GZ(T=0)A ) terms are constrained by conservative theoretical calculation
[95]. All data used in the global fit was corrected for the energy dependence to
the γ − Z box diagram according to the prescription in Ref.[96]. The associated
uncertainties were added in to the systematic error on each data point. The effects
of including points at higher Q2 and θ, as well as varying the mass scale (Λ) of
the dipole parametrization, was studied and the effect were found to be small [36].
Considering the preceding parameters, a standard χ2 minimization is performed
which determines the best fit values for each parameter. The resulting equation is
used to extrapolate to Q2 = 0 allowing for extraction of Qpw(0).
While the global fit includes data as a function of (Q2, θ) it does not lend itself
149
to interpretability when plotting. For this reason the 2-D fit can be rotated into
the forward limit (θ = 0) removing the angular dependence. The rotation was done
using the theoretical calculation of the shift in the asymmetry due to the forward
limit rotation
∆ = Acalc(θ,Q2)− Acalc(0, Q
2). (6.6)
This is then used to shift the measured asymmetry Aep(θ,Q2) into the forward limit.
The reduced asymmetry can then be defined in the following way
Areduce =APV
A= Qp
W +B(θ = 0, Q2)Q2, A =
[(−GFQ
2)
4πα√
2)
](6.7)
This reduced form was used to plot the data points in the global fit as a function of
Q2; the extracted value for QpW is defined as the intercept of the global fit (Q2 → 0).
The extracted values for QpW are shown below. Here the results are presented with
the blinding factor (blinded) and assuming the blinding factor is approximately zero
(Blinding excluded). In the later case the extraction was done without including
the error due to blinding.
QpW (Blindingexcluded) = 0.0845± 0.0093
QpW (Blinded) = 0.0673± 0.0119
QpW (SM) = 0.0705± 0.0008.
(6.8a)
(6.8b)
(6.8c)
The measured result without the blinding factor is ≈ 1.5σ from the SM expectation.
The forward angle global fit of the reduced asymmetry is show in Fig. 6.3 assuming
the blinding factor is very small or zero.
Using the values of the vector couplings in Table. 2.1, the weak-charge can be
written in terms of the quark neutral weak vector couplings as QpW = −2(2Cu1 +
150
FIG. 6.3: The global fit of the reduced asymmetry in the forward limit to Q2 = 0.63(GeV/c)2 is shown as the black line. The red point is the present Qp
W measurement.The yellow band represents the uncertainty of the fit. The dashed blue line showsthe global fit without the measurement of Qp
W presented here. The SM predictionis shown by the arrow. This fit assumes the blinding factor is close to zero.
151
Cd1). Both Cu1 and Cd1 are fit parameters in the global fit used to extract QpW ,
however further constraints can be placed on the neutral weak vector couplings by
considering results from atomic parity-violation experiments. Recent results [97]
from experiments such as [37] on 133Cs provide different linear combinations of the
neutral weak vector couplings that can be used to better separate and determine
Cu1 and Cd1. Combining these results with the PVES results gives values of the
neutral weak vector couplings of
Cu1 = −0.19286± 0.0042 (blindingexcluded)
Cd1 = 0.34373± 0.0041 (blindingexcluded)
Cu1 = −0.1851± 0.0054 (blindingincluded))
Cd1 = 0.3368± 0.0050 (blindingincluded)
The global fit of the quark weak vector couplings can be seen in Fig. 6.4. With
the revised number for couplings, the weak-charge was recalculated according to
QpW = −2(2Cu1 + Cd1). The results of this were found to be
Both the blinded and unblinded values for the calculated QpW match the correspond-
ing values extracted using the reduced asymmetry above. As with the results from
the reduced asymmetry extraction, there is an ∼ 1.5 σ discrepancy between the
measured and SM prediction for the physics asymmetry. These results must be
taken in context though. While the results presented above do correct for the blind-
ing factor, it is still included in the data and must be considered. Presenting the
values without the blinding factor was done to show what the results would be if
152
FIG. 6.4: Constraints on the neutral weak vector constants: isoscalar(Cu1 + Cd1)and isovector (Cu1−Cd1). The APV measurement shown in green mainly constrainsthe isoscalar combination. The blue ellipse includes the current PVES data upto Q2 <0.63 (GeV/c)2 as well as the value reported here. The red ellipse is thecombination of PVES and APV measurements; inner ellipses are 68% confidencelevel and the outer are 95% confidence level. The black line represents the MSscheme SM prediction of sin2 θW [14], with the black dot indicating the SM best fitvalue, sin2 θW = 0.23116.
153
the blinding factor ends up being small. Because of this, drawing scientific conclu-
sions from these results is difficult. The complete analysis of the Run I dataset is
on-going. Following its completion, and subsequent blinding factor removal, more
concrete conclusions will be possible.
6.2 Conclusion & Future Work
While the results presented here represent a factor of two improvement in preci-
sion over the original Wien 0 result, there is still a lot of work to be done. The Run
I data set makes up approximately 33% of the total Qweak data set (in comparison
to the 4% from Wien 0), but completion of the Run II data analysis, which makes
up 63% of the total data set, should see yet more improvement of the statistical
error. The bulk of the systematic error comes from uncertainties on the Aluminium
measurements, Q2, and polarization. Work is currently in progress to improve the
uncertainties of the Aluminium measurement by considering higher order diagrams,
radiative losses, and better understanding of the beamline backgrounds which con-
tribute heavily to the uncertainty of the Aluminium asymmetry. Work on the Q2
measurement is on-going but has been primarily focused on the Run II analysis thus
far. Another source, which has recently been revisited, is the polarization determi-
nation. The issues with the failed coil during Run I have spawned a substantial
amount of simulation and analysis effort by the polarization group. Given that this
is a subtle issue, the effect of which was changing daily during Run I, and man-power
is low, the progress on this has been slow, however, with the shift in focus of the
analysis changing to Run I, more results are expected in the near future.
In terms of work on improving the determination of the beam modulation ex-
tracted detector sensitivities, a substantial amount of work has been done. Many
154
of the issues presented here in the Run I analysis have been rectified. The gaps
in the ramp function have been fixed, the number of independent variables has
been doubled - both the sine and cosine contributions are now considered - and the
BPMs been redefined as “effective” BPMs which have provided much better sepa-
ration among the BPM responses. In general the current analysis efforts have been
focused heavily on the Run II data set and only recently has the focus started to
shift towards Run I; it is expected that this will greatly reduce the systematic errors
as a whole as the analysis ramps up. I have included a more detail write-up with
proposed changes to a future beam modulation system in App. A.4.
155
BIBLIOGRAPHY
[1] Wikimedia Commons. Standard model of elementary particle physics.
[83] J. Hoskins. Beam Polarization Averaging Study. https://qweak.jlab.org/
elog/Analysis+%26+Simulation/1400, 2015.
164
[84] K. A. Aniol et al. Measurement of the neutral weak form factors of the proton.
Phys. Rev. Lett., 82:1096–1100, 1999.
[85] K. A. Aniol et al. Parity-violating electron scattering from 4He and the strange
electric form factor of the nucleon. Phys. Rev. Lett., 96:022003, 2006.
[86] K.A. Aniol et al. Constraints on the nucleon strange form-factors at Q**2 0.1-
GeV**2. Phys.Lett., B635:275–279, 2006.
[87] A. Acha et al. Precision measurements of the nucleon strange form factors at
Q2 ∼ 0.1 gev2. Phys. Rev. Lett., 98:032301, 2007.
[88] Z. Ahmed et al. New precision limit on the strange vector form factors of the
proton. Phys. Rev. Lett., 108:102001, 2012.
[89] D. S. Armstrong et al. Strange-quark contributions to parity-violating asym-
metries in the forward g0 electron-proton scattering experiment. Phys. Rev.
Lett., 95:092001, 2005.
[90] D. Androic et al. Strange quark contributions to parity-violating asymmetries
in the backward angle g0 electron scattering experiment. Phys. Rev. Lett.,
104:012001, 2010.
[91] F. E. Maas et al. Measurement of strange-quark contributions to the nucleon’s
form factors at Q2 = 0.230 (GeV/c)2. Phys. Rev. Lett., 93:022002, 2004.
[92] F. E. Maas et al. Evidence for strange-quark contributions to the nucleon’s
form factors at Q2 = 0.108 (GeV/c)2. Phys. Rev. Lett., 94:152001, 2005.
[93] J. J. Kelly. Simple parametrization of nucleon form factors. Phys. Rev. C,
70:068202, 2004.
165
[94] S. MacEwan. The Weak Charge of the Proton: A Search For Physics Beyond
the Standard Model. PhD. Thesis, University of Manitoba, 2015.
[95] Shi-Lin Zhu, S. J. Puglia, B. R. Holstein, and M. J. Ramsey-Musolf. Nucleon
anapole moment and parity-violating ep scattering. Phys. Rev. D, 62:033008,
2000.
[96] Mikhail Gorchtein, C.J. Horowitz, and Michael J. Ramsey-Musolf. Model-
dependence of the γZ dispersion correction to the parity-violating asymmetry
in elastic ep scattering. Phys.Rev., C84:015502, 2011.
[97] V. A. Dzuba, J. C. Berengut, V. V. Flambaum, and B. Roberts. Revisiting
parity nonconservation in cesium. Phys. Rev. Lett., 109:203003, 2012.
[98] D. Keith Robinson P. R. Bevington. Data Reduction and Error Analysis.
McGraw-Hill, 3rd edition, 2003.
166
APPENDIX A
Appendix
A.1 Detector Error Weighting
The weighted average of a variable x is given by,
〈x〉 =
N∑i
xiwi
N∑i
wi
, (A.1)
where wi is a general weight. The variance of 〈x〉 is given by the error propagation
formula [98],
σ2〈x〉 =
N∑i
(∂〈x〉∂xi
)2
σ2i . (A.2)
167
Applying Eq. A.2 to Eq. A.1,
σ2〈x〉 =
N∑i
wiσiN∑j
wj
2
(A.3)
σ〈x〉 =
√√√√√√√√√√N∑i
wiσiN∑j
wj
2
(A.4)
we obtain the general form of the variance on the weighted average of a variable
x with weight wi. Defining the weighting factor to be the main detector error,
wi = 1/σ2MD we obtain the error on the detector weighted average of x.
σ〈x〉 =
√√√√√√√√√√N∑i
σi
σ2i,MD
N∑j
1
σ2j,MD
2
(A.5)
168
A.2 Rotator Study
FIG. A.1: The survey team used tooling balls that are attached in static locations onthe VDCs to determine position in the frame of the lab. Above shows the positionsand designation of each tooling ball.
During and before the running of Qweak a number of surveys were performed to
analyze the quality of the motion and repeatability of the Region III rotator. The
results of this study were derived from a survey done in September of 2010. The two
main items of interest from the study were the rotational and angular repeatability.
The survey was done using tooling balls which were attached statically to the face
of the VDCs and were used to determine the position with respect to the beamline
in the reference frame of the hall. Placement locations of the tooling balls can be
seen in Fig. A.1.
The numbers shown in Tab. A.1- A.2 are for moving the rotator from home
(defined to be horizontal or at 0) to approximately 45 and back to home. The
listed value is the angle between the ~R(x, y, z) and ~R′(x, y, z) vectors which describe
the radial position before and after rotation. The angle was computed using the
TABLE A.1: Angular and radial repeatability values measured after moving fromHOME→45 → HOME. Radial values are shown in millimeters and angular valuesare in milli-degrees.
TABLE A.2: Angular and radial repeatability values measured after moving fromHOME→45 → HOME. Radial values are shown in millimeters and angular valuesare in milli-degrees.
TABLE A.3: Radial repeatability after moving the chambers from the extended (inbeam) location to the retracted (out of beam) and back. All values are given inmillimeters.
TABLE A.4: Radial repeatability after moving the chambers from the extended (inbeam) location to the retracted (out of beam) and back. All values are given inmillimeters.
Law of Cosines for ~R and ~R′. The Radial repeat is defined as δr = ~R − ~R′ or the
difference in the radial vectors before and after rotation. This details how static
the radial position of the VDC was after a set of rotations. Some values which are
missing were not measured for both VDC packages due to time constraints.
Tables A.3 adn A.4 are the difference in the measured radius of each tooling
ball after the rotator has been retracted and re-extended. You will notice a signif-
icant difference in the size of the δx value for the measurement of Vader. We have
TABLE A.12: Shown is the extracted angle with respect to the home position.
Vader radial repeatability after rotation to the -90 positionPoint Angle Repeat Radial RepeatVAH 8.216 0.799VBH 8.021 0.736VCH 7.786 0.673VDH 2.375 1.131VEH 2.380 1.132VFH 7.792 0.672VGH 8.092 0.737VHH 8.328 0.806
TABLE A.13: The repeatability of the measured angle and radius of the chambersafter repeated rotation to the -90 position. Angles are shown in milli-degress andradial values are shown in millimeters.
Yoda radial repeatability after rotation to the -90 positionPoint Angle Repeat Radial RepeatYAH 8.111 0.799YBH 6.898 0.736YCH 7.749 0.673YDH 11.358 1.131YEH 12.582 1.132YFH 6.890 0.672YGH 7.587 0.737YHH 7.761 0.807
TABLE A.14: The repeatability of the measured angle and radius of the chambersafter repeated rotation to the -90 position. Angles are shown in milli-degress andradial values are shown in millimeters.
175
Standard Deviation Radial Measurement in Theta: VaderPoint Radial σ(x, y, z) Radial σ(x, y)VAH 1.577 3.473VBH 1.619 3.246VCH 1.705 3.482VDH 1.597 3.769VEH 1.596 3.789VFH 1.706 3.534VGH 1.643 3.540VHH 1.582 3.664
TABLE A.15: The standard deviation of the radial measurement for each point ineach position in theta. Extracted values are shown for the radius in both the (x,y)plane and the (x,y,z) plane. Values in mm.
Standard Deviation Radial Measurement in Theta: YodaPoint Radial σ(x, y, z) Radial σ(x, y)VAH 1.678 2.468VBH 1.820 2.734VCH 1.559 2.475VDH 1.883 2.967VEH 1.579 2.479VFH 1.902 2.979VGH 1.841 2.506VHH 1.983 2.795
TABLE A.16: The standard deviation of the radial measurement for each point ineach position in theta. Extracted values are shown for the radius in both the (x,y)plane and the (x,y,z) plane. Values in mm
176
Standard Deviation of Angular Measurements: VaderAngle σ(θ)-45 0.021-90 0.00945 0.02090 0.027
TABLE A.17: The standard deviation of the angles measured with respect to homeof each measurement shown in Tables A.5-A.11.
Standard Deviation of Angular Measurements: YodaAngle σθ-45 N/A-90 0.02145 0.01690 0.015
TABLE A.18: The standard deviation of the angles measured with respect to homeof each measurement shown in Tables A.6-A.12.
The standard deviation of the measured angle between each tooling point be-
tween home and each rotated position (-90, -45, 0, 45, 90) is shown in Tables
A.17 and A.18. All units are in degrees and show that the angle between octants is
within milli-degrees of the ideal separation.
The point of the measurements found in Tables A.19 and A.20 was to check
for significant chamber sagging/flexing and bound the resolution of four measure-
ments. For each rotational position the distance between pairs of tooling points
is computed. As the chambers are static objects these points should not change
within the resolution of the measurement. This gives an idea how well the position
of each point was determined or the effective resolution considering any flexing or
sagging effect that might be present. The distance between each point pair is given
as well as the standard deviation between rotational positions. All units are in mm.
From these tables we can determine that the point-to-point resolution of the survey
measurements is on the few micron level.
177
Distance between static points: VaderPoint Pairs BA DF CE HG0 412.490 839.544 839.825 412.379-45 412.475 839.534 839.825 412.382-90 412.488 839.540 839.818 412.38345 412.482 839.536 839.823 412.38690 412.477 839.545 839.819 412.387
σ 0.007 0.005 0.003 0.003
TABLE A.19: For each rotational position the distance between pairs of toolingball locations (see Fig. A.1) is calculated. As these are static points the distancebetween them should not change. This gives a good estimate of the resolution withwhich the points were measured.
Distance between static points: YodaPoint Pairs BA DF CE HG0 412.499 292.063 291.913 412.414-45 N/A N/A N/A N/A-90 412.510 292.060 291.900 412.40545 412.501 292.065 291.917 412.414390 412.497 292.063 291.903 412.415
σ 0.006 0.002 0.008 0.005
TABLE A.20: For each rotational position the distance between pairs of toolingball locations (see Fig. A.1) is calculated. As these are static points the distancebetween them should not change. This gives a good estimate of the resolution withwhich the points were measured.
178
A.3 Background Correction Error Propagation
The calculation of the individual systematic error contributions from the back-
ground correction to the physics asymmetry is done according to
σ2A =
∣∣∣∣∂A∂b∣∣∣∣2 σ2
b . (A.7)
Applying this to Eq. 6.1 we get the following relations:
∂A|Amsr =R× dAmsr
P (1− ftotal), (A.8a)
∂A|Ablind=R× dAblinding
msr
P (1− ftotal), (A.8b)
∂A|P =−R× Amsr
(1− ftotal)
dP
P 2, (A.8c)
∂A|Ai=−R× fi
(1− ftotal)dAi (A.8d)
∂A|R = dR×(Amsr
P− ΣifiAi
)(1− ftotal)
(A.8e)
∂A|fi = εijkl
(Amsr
P+ (fj + fk + fl)Ai − fjAj − fkAk − flAl
)R× dfi
(1− ftotal)
(A.8f)
In the final equation the Levi-Civita operator is invoked to denote the permuta-
tion of the indices 1→2→3→4. With these calculations, the total systematic error
contribution from backgrounds and experimental bias is given by
σA =√
(∂A|Amsr)2 + (∂A|P )2 + (∂A|Ai
)2 + (∂A|R)2 + (∂A|fi)2. (A.9)
179
A.4 Beam Modulation Improvements
As was laid out in the Beam Modulation chapter, the system as a whole was
a success in measurement of the helicity-correlated beam sensitivities. With every
success however there is room for improvement, and there were a number of problems
that were discovered in the system, that in hindsight, could have been fixed. In the
interest of future modulation systems that might be built, I will address some of
those issues and ways that they could have been avoided.
Fast-feedback This was a big issue in the analysis and running of the beam modu-
lation system. By leaving the fast-feedback system on during modulation another
level of complexity was added into the analysis; because the fast-feedback system
worked to counteract modulations in the beam position and angle the total re-
sponse became a composite function with a phase out of response to the driving
signal. This issue was missed in early testing of the system but in future iter-
ations fast-feedback should be paused for the duration of the beam modulation
period. This will lead to cleaner, more interpretable results.
Diagnostics and Tuning One of the pressing issues during the design, construc-
tion, and testing of the modulation system was a lack of manpower to provide
the tools and do the diagnostic testing that was required to make the system the
best it could be. A robust system would include online monitoring and diagnos-
tic tools to give run-by-run results for things such as: beam monitor responses,
detector responses, extracted sensitivities, and beam trajectory plots. The later
of these would be enormously helpful as it would enable shift crew to monitor
how separated the modulations were in phase space and call in experts to tune
the modulation coils, given changes in the beam properties. This would ensure
that the modulations - for example X-position and X-angle - were as pure posi-
180
tion and angle as possible at all times leading to better extracted results. Too
many times during the experimental running we didn’t find problems until days
after they appeared. A robust, real-time diagnostic system would ensure better
response time to problems, and better results with each measurement. This of
course requires an adequate amount of manpower to design and implement the